Power distribution network fault positioning method and system based on multi-source sensing fusion

Abstract

The invention discloses a power distribution network fault positioning method and system based on multi-source sensing fusion, and relates to the technical field of intelligent power distribution, and the method comprises the following steps: collecting high-frequency fluctuation signals corresponding to a plurality of sensing points around the transient high-frequency disturbance generated at the initial stage of a power distribution network fault, expanding each high-frequency fluctuation signal under a unified time reference, and obtaining a plurality of high-frequency fluctuation signals; and extracting starting and ending moments of protruding segments in the high-frequency fluctuation signal, and forming a local amplification fingerprint set based on the protruding segments. According to the invention, by introducing local amplification fingerprint identification and propagation consistency constraint, the pseudo propagation influence is eliminated, so that the fault trace is more consistent with the real propagation law, and the positioning accuracy and stability are improved; and meanwhile, through bandwidth convergence type peak clipping and energy boundary retention, in combination with time and space anchoring of a main fault trace, accurate locking and dynamic isolation are realized, and the fault response speed and the system reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent power distribution technology, specifically to a method and system for fault location in power distribution networks based on multi-source sensor fusion. Background Technology

[0002] Multi-source sensor fusion-based fault location in distribution networks involves simultaneously collecting operational information from different locations and types of sensors during network operation. This information is then uniformly aligned, correlated, and comprehensively analyzed to pinpoint the specific section or node where the fault occurred. The core idea is to deploy multiple sensing devices at key locations such as distribution lines, switching equipment, and transformer nodes to continuously acquire operational information such as voltage changes, current fluctuations, switch operation status, and transient anomalies. By cross-verifying multi-source data in time and space, the problems of susceptibility to interference and misjudgment associated with single signals are eliminated, making fault characteristics clearer and more identifiable. Building upon this, fault prediction and health management mechanisms are introduced to continuously analyze trends and assess the health of the sensed operational data. By identifying early anomalies and signs of aging in key equipment, potential fault risks are warned in advance. This expands the approach from passive fault location to proactive prediction and full lifecycle health management, shifting the distribution network's operational status from post-event identification to pre-event prevention, further enhancing the safety and stability of the distribution network.

[0003] The existing technology has the following shortcomings: Under current technological conditions, fault location in distribution networks typically relies on the time alignment and joint analysis of multi-point sensor information to determine the direction of fault propagation. In the early stages of a fault, transient high-frequency disturbances of extremely short duration inevitably occur in the line. When these disturbances are abnormally amplified by sensors at local locations and used as key features in the fusion process, existing technologies, during multi-source information time alignment, easily misidentify these amplified signals as the starting point of fault propagation along the line. This leads to the construction of a non-existent fault propagation path in the fusion results. Consequently, the fault location results are biased towards sections far from the actual fault point, causing the system to prioritize isolating non-faulty sections, while the actual faulty section fails to be isolated immediately. This results in the continuous transmission and accumulation of fault current and abnormal energy in the main line, significantly amplifying the operational risks of the distribution network.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for fault location in power distribution networks based on multi-source sensor fusion, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a fault location method for distribution networks based on multi-source sensor fusion, comprising the following steps: To address the transient high-frequency disturbances generated in the early stages of a distribution network fault, high-frequency fluctuation signals corresponding to multiple sensor points are collected. Under a unified time reference, each high-frequency fluctuation signal is expanded, and the start and end times of the spike segments in the high-frequency fluctuation signals are extracted. Based on the spike segments, a locally magnified fingerprint set is formed. Based on the local magnified fingerprint set, a propagation consistency constraint relationship is constructed. The arrival order of each protruding segment in the local magnified fingerprint set is bound to the corresponding distribution network topology distance. Protruding segments whose arrival order deviates from the distribution network topology distance are identified, forming a list of suspected false propagation points. Based on the list of suspected false propagation points, bandwidth convergence peak clipping is performed on the protruding segments corresponding to the list of suspected false propagation points, while retaining the energy transition boundary information, to form the main fault trace after removing the local amplification effect. Based on the main fault trace, the earliest stable inflection point is determined in the time dimension, and the shortest consistent segment is mapped in the spatial dimension. Based on the earliest stable inflection point and the shortest consistent segment, a real fault segment locking list is generated. Based on the real fault section locking list, an extremely short silent time window is introduced within the real fault section to suppress local amplification triggering, and the isolation command timing is dynamically guided according to the earliest stable inflection point of the main fault trajectory, and the isolation action unfolds along the main fault trajectory.

[0007] Preferably, the steps for forming a locally magnified fingerprint set are as follows: Multiple sensing devices are deployed at the locations of power distribution lines, switching equipment and transformer nodes to collect high-frequency fluctuation signals and to unfold each high-frequency fluctuation signal with a unified time base. The unfolded high-frequency wave signal is continuously scanned in time to extract the spike segments with increased amplitude change rate in the high-frequency wave signal, and the start and end times of the spike segments are determined. The protrusion segments of each sensing point are cross-mapped under a unified time base, and the amplitude change pattern and duration distribution of the protrusion segments are analyzed to form a locally magnified fingerprint unit. The magnified fingerprint units of each sensor point are summarized according to time sequence and spatial location to form a magnified fingerprint set, which is used for subsequent propagation discrimination and fault trace extraction.

[0008] The preferred method for forming the list of suspicious points related to false propagation is as follows: After obtaining the locally magnified fingerprint set, the arrival time of the protruding segments in the locally magnified fingerprint set is mapped to the global time axis under a unified time base, and the spatial location mapping relationship of each sensing point in the distribution network topology is established. An arrival sequence is constructed based on the temporal relationship between multiple protruding segments, and the arrival sequence is bound to the distribution network topology distance level one by one; A correlation analysis was performed on the binding results between the arrival order and the distribution network topology distance to identify abrupt segments where the arrival order and the distribution network topology distance showed a reverse offset. The protruding segments that show reverse offset are collected to form a list of suspected false propagation points, which are used for subsequent stripping away of local magnification effects and extraction of the main fault trace.

[0009] Preferably, when forming the list of suspected pseudo-propagation points, the temporal attributes, topological distance attributes, and locally magnified fingerprint features of the protruding segments are retained, and the temporal continuity and spatial consistency of the protruding segments are maintained under a unified time benchmark, so that the list of suspected pseudo-propagation points can accurately reflect the distribution characteristics of non-real propagation segments and provide a stable data foundation for subsequent high-frequency feature stripping.

[0010] Preferably, the steps for forming the main fault trace based on the list of suspected false propagation points are as follows: After obtaining the list of suspected false propagation points, the high-frequency fluctuation signals corresponding to the protruding segments in the list of suspected false propagation points are mapped to the time axis under a unified time base, while maintaining the start and end time boundaries of the protruding segments. Based on the mapped protruding segments, bandwidth convergent peak clipping is performed on the protruding segments, and the peak clipping is limited to the start and end time range of the protruding segments. During the bandwidth convergence peak-shaving process, the energy transition boundary information in the spike segments is preserved. The spike segments that have undergone peak shaving and retain energy transition boundary information are integrated with the remaining spike segments according to a unified time reference to form the main fault trace.

[0011] Preferably, during the formation of the main fault trace, the bandwidth convergent peak clipping process uses the energy distribution state of the protruding segment as a constraint condition to suppress the high-frequency energy peaks inside the protruding segment, while maintaining the continuous existence of the energy transition boundaries corresponding to the start and end positions of the protruding segment, so that the main fault trace remains completely connected in the time dimension and avoids interference from local amplification features in the energy dimension to the fault propagation judgment.

[0012] Preferably, the steps for generating a list of actual faulty sections based on the main fault trace are as follows: After the main fault trace is formed, it is unfolded under a unified time base, and the changing trend of the main fault trace is analyzed along the time axis to determine the earliest stable inflection point of continuous change from a stable state to an abnormal state. Based on the earliest stable inflection point, the main fault trace is mapped to the distribution network topology in the spatial dimension, and the distribution of the main fault trace at different spatial locations is unfolded. Based on spatial expansion, the continuous distribution range of the main fault trace in the distribution network topology is compressed and mapped to extract the shortest consistent segment corresponding to the earliest stable inflection point. By combining the earliest stable inflection point and the shortest consistent segment, a list of real fault segments is generated for subsequent control and processing.

[0013] Preferably, in the process of generating the real fault section locking list, the continuous distribution relationship of the main fault trace in the distribution network topology is used as a constraint condition to ensure that the shortest consistent section is continuously mapped in the spatial dimension, and that the earliest stable inflection point is associated with the shortest consistent section in the time dimension, thereby limiting the spatial range and time orientation of the real fault section locking list to be consistent.

[0014] Preferably, the isolation action steps based on the actual faulty section lockout list are as follows: After obtaining the list of real faulty sections, the faulty section information in the list is matched with the time distribution information of the main fault trace, and an extremely short silent time window is introduced at the starting position of the real faulty section. After the extremely short silent time window ends, the timing of isolation commands is dynamically adjusted based on the earliest stable inflection point of the main fault trace. Based on the time rhythm after dynamic traction, the isolation action is carried out sequentially within the actual fault section along the propagation direction of the main fault trajectory. During the isolation process, the isolation actions of each section within the actual fault section are sequentially linked in time to form a sequence of isolation actions that unfolds continuously along the main fault trajectory.

[0015] A distribution network fault location system based on multi-source sensor fusion includes a high-frequency disturbance recording module, a propagation consistency discrimination module, a high-frequency feature stripping module, a dual-axis fault location module, and an isolation action control module. The high-frequency disturbance recording module collects high-frequency fluctuation signals corresponding to multiple sensor points based on transient high-frequency disturbances generated in the early stage of distribution network faults. It expands each high-frequency fluctuation signal under a unified time base, extracts the start and end times of the protruding segments in the high-frequency fluctuation signals, and forms a locally magnified fingerprint set based on the protruding segments. The propagation consistency discrimination module constructs a propagation consistency constraint relationship based on the local magnified fingerprint set. It binds the arrival order of each protruding segment in the local magnified fingerprint set to the corresponding distribution network topology distance, identifies protruding segments whose arrival order deviates from the distribution network topology distance, and forms a list of suspected false propagation points. The high-frequency feature stripping module, based on the list of suspected pseudo-propagation points, performs bandwidth convergence-type peak clipping on the protruding segments corresponding to the list of suspected pseudo-propagation points, while retaining the energy transition boundary information, forming the main fault trace after removing the local amplification effect. The dual-axis fault location module determines the earliest stable inflection point in the time dimension and maps the shortest consistent segment in the spatial dimension based on the main fault trajectory. It generates a real fault segment lock list based on the earliest stable inflection point and the shortest consistent segment. The isolation action control module, based on the real fault section locking list, introduces an extremely short silent time window within the real fault section to suppress local amplification triggering, and dynamically guides the timing of isolation commands according to the earliest stable inflection point of the main fault trajectory, so that the isolation action unfolds along the main fault trajectory.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention introduces local magnification fingerprint recognition and propagation consistency constraints during multi-source sensor fusion, establishing a unified constraint relationship between the temporal sequence of fault propagation characteristics and the distribution network topology distance. This effectively identifies and eliminates false propagation information caused by local magnification effects in the early stages of a fault. Through this method, the formation of fault propagation traces more closely matches the actual physical propagation patterns of the line, avoiding misjudgments during the time alignment stage, improving the accuracy and stability of fault location, and making the judgment results of fault propagation paths more consistent and reliable.

[0017] This invention achieves precise fault segment localization and dynamic matching of isolation actions by applying bandwidth-convergent peak-shaving processing to spurious propagation segments while preserving energy transition boundary information, combined with the earliest stable inflection point and shortest consistent segment of the main fault trajectory. This technique ensures the continuity of fault propagation characteristics while suppressing local amplification triggering, ensuring that isolation actions strictly follow the main fault trajectory. This shortens fault response time, reduces the probability of erroneous actions, and improves the timing coordination of fault isolation and the safety and reliability of system operation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method for fault location in power distribution networks based on multi-source sensor fusion according to the present invention.

[0020] Figure 2 This is a schematic diagram of the modules of the power distribution network fault location system based on multi-source sensor fusion of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] This invention provides, for example Figure 1 The distribution network fault location method based on multi-source sensor fusion shown includes the following steps: To address the transient high-frequency disturbances generated in the early stages of a distribution network fault, high-frequency fluctuation signals corresponding to multiple sensor points are collected. Under a unified time reference, each high-frequency fluctuation signal is expanded, and the start and end times of the spike segments in the high-frequency fluctuation signals are extracted. Based on the spike segments, a locally magnified fingerprint set is formed. To address the transient high-frequency disturbances generated in the initial stage of a distribution network fault, high-frequency fluctuation signals from multiple sensing points are expanded and feature extracted under a unified time reference to form a locally magnified fingerprint set for subsequent fault propagation identification and location. The specific steps are as follows: Multiple sensors are deployed at key locations such as power distribution lines, switchgear, and transformer nodes to collect high-frequency fluctuation signals. During the initial transient process of a fault in the power distribution network, each sensor collects the high-frequency components of the voltage and current waveforms at its corresponding monitoring point in real time. To ensure the comparability and synchronization of data collected from different sensor points, all sensors are calibrated using a unified time reference. This unified time reference originates from a high-precision clock signal distribution device, allowing high-frequency fluctuation signals from different sensor points to be displayed on the same time coordinate system. This unified time reference enables the high-frequency fluctuation signals from each sensor point to be arranged correspondingly on the same time axis, providing a complete timing basis for the subsequent extraction of spike segments.

[0023] After completing the time-uniform unfolding of the high-frequency fluctuation signals, a continuous time-series scan is performed on the high-frequency fluctuation signals corresponding to each sensing point to identify fluctuation segments with a significant increase in amplitude change rate within a short time interval. These fluctuation segments are defined as spike segments. During the identification of spike segments, the start and end times of the spike segments are determined by comparing the amplitude change trend of each fluctuation signal time-by-time, ensuring that each spike segment corresponds to a complete high-frequency disturbance duration interval. To avoid time misjudgment caused by transient glitches, this step compares and calibrates the spike segments of adjacent sensing points on a unified time axis, ensuring that the time boundaries of the spike segments are consistent and continuous at a unified time scale. After this processing, the high-frequency fluctuation signals of each sensing point form a set of spike segments with clear time boundaries, laying the foundation for the subsequent formation of a locally magnified fingerprint set.

[0024] After obtaining the set of protrusion segments from each sensor point, these segments are cross-mapped under a unified time reference to analyze the amplitude variation and duration distribution of protrusion segments from different sensor points within the same time interval. This cross-mapping method reveals that certain sensor points exhibit significant amplification characteristics within the same time interval, characterized by large amplitude increases, concentrated energy, steep rise edges, and short durations. These characteristics are defined as local amplification features. In this step, these local amplification features are correlated one-to-one with the start and end times of the protrusion segments, the amplitude variation magnitude, and the corresponding spatial location of the sensor points, forming local amplification fingerprint units containing multi-dimensional information. Each local amplification fingerprint unit corresponds to a sensor point protrusion segment with a characteristic fluctuation pattern. In this way, the originally continuous high-frequency fluctuation signal is transformed into a fingerprint-like feature set that can be compared and identified, enabling rapid identification of potentially misjudged local amplification responses in subsequent fault propagation analysis.

[0025] After extracting the local magnified fingerprint units, all the local magnified fingerprint units corresponding to all sensing points are summarized according to time sequence and spatial location to form a complete local magnified fingerprint set. The local magnified fingerprint set not only includes the start and end times, amplitude variation characteristics, and energy concentration of each sensing point's protrusion segment, but also spatial distribution information corresponding to the distribution line topology. This spatiotemporal fusion summarization method ensures that the local magnified fingerprint set retains the independent high-frequency disturbance characteristics of each measuring point while achieving a unified representation of multi-point signals in the distribution network on a global time coordinate. During the formation of the local magnified fingerprint set, the time boundary, energy transition characteristics, and spatial location of each protrusion segment are consistently mapped to ensure comparability and continuity among the fingerprint units within the local magnified fingerprint set, thus providing a complete, accurate, and traceable data foundation for subsequent propagation consistency discrimination and fault trace extraction.

[0026] Based on the local magnified fingerprint set, a propagation consistency constraint relationship is constructed. The arrival order of each protruding segment in the local magnified fingerprint set is bound to the corresponding distribution network topology distance. Protruding segments whose arrival order deviates from the distribution network topology distance are identified, forming a list of suspected false propagation points. Based on the existing locally magnified fingerprint set, a correlation analysis is conducted between the temporal sequence of each protruding segment in the locally magnified fingerprint set and the physical topology of the distribution network by introducing propagation consistency constraints. This aims to identify non-genuine propagation phenomena caused by the local magnification effect and generate a list of suspected pseudo-propagation points, providing a basis for subsequent high-frequency feature stripping and fault trace extraction. The specific steps are as follows: After obtaining the locally magnified fingerprint set, the time series of each protruding segment contained in the set is processed, and the arrival time of each protruding segment is uniformly mapped to the global time axis under the aforementioned unified time reference. Simultaneously, based on the actual physical connection relationship of the distribution network lines, a spatial location mapping relationship between each sensor point and the distribution network topology is established, and the distance level of the sensor point corresponding to each protruding segment within the distribution network topology is labeled. In this way, each protruding segment in the locally magnified fingerprint set possesses both a clear temporal location attribute and a clear topological distance attribute, thus providing complete temporal-spatial foundational information for the subsequent construction of propagation consistency relationships.

[0027] After mapping the temporal attributes of the spike fragments to their topological distance attributes, an arrival sequence of the spike fragments is constructed based on the temporal relationship between multiple spike fragments within the same fault event window. This arrival sequence reflects the arrangement of each spike fragment on a unified time axis. Simultaneously, this arrival sequence is bound one-to-one with the corresponding distribution network topological distance level, establishing a one-to-one correspondence between the arrival order of each spike fragment and its physical topological distance. This sequential binding method provides a clear indication of the propagation trend of high-frequency disturbances in the distribution network topology, laying the foundation for determining whether the propagation path conforms to physical propagation logic.

[0028] After binding the arrival order with the distribution network topology distance, a comprehensive correlation analysis is performed on the binding results to identify abrupt segments where the arrival order and topology distance are inconsistent. Specifically, when a abrupt segment arrives earlier in the time series but its corresponding distribution network topology distance is at a relatively far level, or when a abrupt segment arrives later in the time series but its corresponding distribution network topology distance is at a relatively close level, it is determined that there is a reverse offset relationship between the arrival order of the abrupt segment and the distribution network topology distance. By continuously identifying this reverse offset relationship, abrupt segments that conform to the physical propagation logic and those that deviate from it can be distinguished, thereby separating non-true propagation segments that may be caused by local amplification effects from the normal propagation sequence.

[0029] After identifying abrupt segments whose arrival order deviates inversely from the distribution network topology distance, these segments are individually aggregated from the local magnification fingerprint set to form a list of suspected pseudo-propagation points. Each abrupt segment in the list retains its original temporal attributes, topological distance attributes, and corresponding local magnification fingerprint features, enabling targeted processing of these segments in subsequent steps. By creating this list, potential non-propagation information can be centrally managed without disrupting the overall local magnification fingerprint set structure. This provides clear and traceable processing targets for subsequent removal of local magnification effects and extraction of the main fault trace.

[0030] Based on the list of suspected false propagation points, bandwidth convergence peak clipping is performed on the protruding segments corresponding to the list of suspected false propagation points, while retaining the energy transition boundary information, to form the main fault trace after removing the local amplification effect. Based on the existing list of suspected false propagation points, targeted bandwidth-converging peak-shaving processing is applied to the corresponding spike segments in the list, while fully preserving energy transition boundary information during the peak-shaving process. This eliminates the interference of local amplification effects on fault propagation judgment, forming a main fault trace that truly reflects fault propagation behavior. The specific steps are as follows: After obtaining the list of suspected false propagation points, each protruding segment in the list is expanded one by one. The high-frequency fluctuation signal corresponding to each protruding segment is remapped onto a time axis under a unified time reference, while maintaining its original start and end time boundaries. In this way, each protruding segment in the list of suspected false propagation points forms a continuous connection with the previous steps in the time dimension. At the same time, it ensures that the subsequent processing only targets the protruding segments defined by the list of suspected false propagation points, without affecting other protruding segments in the locally magnified fingerprint set that are not included in the list of suspected false propagation points, thus ensuring the targeting and continuity of the processing scope.

[0031] After completing the time unfolding of the corresponding spikes in the list of suspected false propagation points, bandwidth convergence-based peak clipping is performed around these spikes. Specifically, based on the spectral distribution characteristics of the high-frequency fluctuation signals within the spikes, the effective bandwidth of the high-frequency fluctuation signals is gradually converged, suppressing excessively concentrated high-frequency energy peaks during the time unfolding process, while non-concentrated energy regions maintain their continuous variation. During bandwidth convergence-based peak clipping, the start and end times of the spikes are always used as processing boundaries to ensure that the peak clipping effect only occurs within the spikes, thus avoiding impact on normal fluctuation regions outside the spikes. This gradual bandwidth convergence-based peak clipping method effectively weakens abnormal peaks caused by local amplification effects while maintaining the overall temporal structure of fault-related fluctuations.

[0032] While implementing bandwidth-convergent peak-shaving processing, the energy transition boundary information within the spike fragments is preserved. This energy transition boundary information represents the boundary positions where high-frequency fluctuation signals transition from a stable state to an abnormal state, or from an abnormal state back to a stable state, in the time dimension. It is a crucial temporal characteristic reflecting the occurrence and propagation of faults. In this step, while weakening the abnormal peak values ​​within the spike fragments, the energy change trends at the start and end boundaries of the spike fragments are not weakened, ensuring that the boundary contours of the spike fragments remain clearly discernible on the time axis. In this way, even if the local amplification effect is suppressed, the fault-related energy transition positions corresponding to the spike fragments are completely preserved, thus providing a reliable temporal anchoring basis for subsequently extracting the true fault propagation trajectory.

[0033] After bandwidth convergence-based peak-shaving processing of the corresponding spike segments in the pseudo-propagation suspicion list while retaining energy transition boundary information, the processed spike segments are integrated with other spike segments not included in the pseudo-propagation suspicion list and rearranged according to a unified time base to form a continuous fault propagation trajectory after removing the effects of local amplification. This continuous fault propagation trajectory reflects the true propagation time sequence of each high-frequency disturbance in the distribution network after suppressing the local amplification effect, constituting the main fault trajectory. The main fault trajectory maintains the continuity between spike segments in the time dimension and avoids the interference of abnormal peaks on the overall shape in the energy dimension, thus more accurately reflecting the true process of fault occurrence and its spread along the distribution network topology.

[0034] Based on the main fault trace, the earliest stable inflection point is determined in the time dimension, and the shortest consistent segment is mapped in the spatial dimension. Based on the earliest stable inflection point and the shortest consistent segment, a real fault segment locking list is generated. Based on the established main fault trace after removing the effects of local amplification, a joint analysis of the main fault trace in both time and spatial dimensions is performed to determine a time anchor point that stably reflects the true initiation characteristics of the fault. Then, a spatial range highly consistent with this time anchor point is locked within the distribution network topology, thereby generating a list of locked true fault sections. This provides clear and reliable spatial objects for subsequent control processes. The specific steps are as follows: After obtaining the main fault trace, it is unfolded as a whole under a unified time reference, and its changing trend is continuously observed along the time axis. By segmenting and analyzing the morphological changes of the main fault trace in the time dimension, the focus is on the region where the main fault trace transitions from a stable state to an abnormal state, identifying the time inflection points that can maintain a stable changing trend. These time inflection points are different from the brief fluctuations caused by transient fluctuations; rather, they represent the starting points where the main fault trace continuously exhibits the same direction of change over a certain time span. In this way, the earliest stable inflection point reflecting the true starting point of the fault is extracted from the main fault trace in the time dimension, so that the subsequent localization process is no longer affected by early transient disturbances.

[0035] After determining the earliest stable inflection point, this point is used as the time anchor to analyze the spatial distribution of the main fault trace. Specifically, based on the performance of the main fault trace at each sensor point at its corresponding time position, the states of each sensor point corresponding to the earliest stable inflection point time are mapped to the distribution network topology, thereby observing the coverage of the main fault trace at different spatial locations within the physical connectivity of the distribution network. In this process, the focus is on analyzing the continuity and consistency distribution characteristics of the main fault trace in the distribution network topology before and after the earliest stable inflection point time, providing a spatial basis for subsequently determining the shortest consistent segment.

[0036] After completing the spatial unfolding corresponding to the earliest stable inflection point, the spatial range in the distribution network topology that is highly consistent with the main fault trace is compressed and mapped, gradually reducing the coverage area of ​​the main fault trace in the spatial dimension. During the compression process, the continuous performance of the main fault trace between adjacent topology nodes is used as a constraint condition to eliminate spatial locations that only appear briefly at individual nodes and do not form a continuous expansion relationship, thus retaining segments in the topology that exhibit continuous and unidirectional change characteristics. In this way, the shortest consistent segment that is highly consistent with the earliest stable inflection point time is extracted from the main fault trace in the spatial dimension, so that this segment can truly reflect the actual impact range of the fault in the distribution network.

[0037] After simultaneously obtaining the earliest stable inflection point and the shortest consistent segment, the earliest stable inflection point is used as the time anchoring condition, and the shortest consistent segment as the spatial locking condition. These two conditions are then combined to form a real fault segment locking list. This list records the time information corresponding to the earliest stable inflection point, as well as the range of the shortest consistent segment locked within the distribution network topology, ensuring clear directionality in both time and space for fault location. By generating this real fault segment locking list, the affected area can be precisely limited to a segment highly consistent with the main fault trajectory without expanding the location range, providing a clear and reliable basis for subsequent control operations targeting that segment.

[0038] Based on the real fault section locking list, an extremely short silent time window is introduced in the real fault section to suppress local amplification triggering, and the isolation command timing is dynamically guided according to the earliest stable inflection point of the main fault trajectory, and the isolation action unfolds along the main fault trajectory. Based on the existing list of real faulty sections, an extremely short silent time window is introduced within each real faulty section to suppress local amplification triggering. Then, the timing of isolation commands is dynamically guided according to the earliest stable inflection point of the main fault trajectory, ensuring that the isolation actions strictly follow the temporal evolution sequence of the main fault trajectory. This achieves precise isolation and timely removal of the faulty sections. The specific steps are as follows: After obtaining the list of locked actual fault sections, the fault section information in the list is correlated with the time distribution information of the main fault trace to determine the time-dimensional variation range of the main fault trace for each locked fault section. After establishing the correspondence, an extremely short silent time window is introduced at the beginning of each actual fault section. The purpose of this extremely short silent time window is to ensure that the protection devices, switching devices, and sensors within the fault section maintain monitoring without triggering actions for a very short period when the fault section enters the isolation preparation phase, thereby suppressing erroneous triggering caused by locally amplified signals or transient fluctuations. In this way, a time buffer can be established before the isolation action begins, allowing local disturbance signals to naturally attenuate within the silent time window, thus providing a stable environment for the accurate triggering of subsequent isolation commands.

[0039] After the extremely short silent time window ends, the timing of isolation commands is dynamically adjusted based on the earliest stable inflection point identified in the main fault trajectory. Specifically, the earliest stable inflection point serves as the time anchor point for fault propagation, reflecting the critical time position from fault occurrence to stable expansion. During the isolation command triggering process, by aligning the time sequence of isolation commands with the earliest stable inflection point of the main fault trajectory, the initiation time of the isolation action is synchronized with the temporal evolution of fault propagation. This ensures that the isolation action unfolds smoothly when the main fault trajectory reaches a stable stage, thereby avoiding energy backlash or protection malfunctions caused by premature or delayed actions. In this step, the core function of dynamic adjustment is to ensure that the rhythm of the isolation action is coordinated with the changing rhythm of the main fault trajectory, making the isolation behavior a response to the natural continuation of fault propagation.

[0040] After dynamically guiding the isolation command timing, the isolation action is gradually deployed along the propagation direction of the main fault trajectory. During this deployment, following the spatial extension sequence of the main fault trajectory in the distribution network topology, starting from the node corresponding to the earliest stable inflection point, switching devices located within the actual fault section are triggered sequentially, causing the isolation action to expand outward along the fault propagation direction. The action time of each switching device is consistent with the energy change time of the main fault trajectory at the corresponding spatial location, ensuring that the timing of the isolation action strictly matches the propagation law of the main fault trajectory. Through this trajectory-based deployment method, the isolation behavior of the distribution network is highly consistent with the fault development process, thereby achieving the gradual disconnection of fault energy, preventing non-faulty sections from being prematurely isolated, and ensuring that the actual faulty section is effectively isolated in the first instance.

[0041] During the isolation process along the main fault trajectory, a time-series connection is performed on all isolated segments to form a continuous response chain of isolation actions between different segments. Specifically, after the isolation action of the previous level segment is completed, the isolation action of the next level segment unfolds sequentially in time according to the propagation order of the main fault trajectory, thus forming a coherent sequence of isolation actions across the entire real fault segment. Through this continuous isolation method, the fault energy is gradually blocked and dissipated along the main fault trajectory, achieving a complete isolation loop from the time dimension to the spatial dimension. After isolation is completed, the status information of each segment recorded in the real fault segment lock list can be updated to the isolation completed state, providing clear boundary conditions for subsequent fault recovery and reconstruction.

[0042] This invention introduces local magnification fingerprint recognition and propagation consistency constraints during multi-source sensor fusion, establishing a unified constraint relationship between the temporal sequence of fault propagation characteristics and the distribution network topology distance. This effectively identifies and eliminates false propagation information caused by local magnification effects in the early stages of a fault. Through this method, the formation of fault propagation traces more closely matches the actual physical propagation patterns of the line, avoiding misjudgments during the time alignment stage, improving the accuracy and stability of fault location, and making the judgment results of fault propagation paths more consistent and reliable.

[0043] This invention achieves precise fault segment localization and dynamic matching of isolation actions by applying bandwidth-convergent peak-shaving processing to spurious propagation segments while preserving energy transition boundary information, combined with the earliest stable inflection point and shortest consistent segment of the main fault trajectory. This technique ensures the continuity of fault propagation characteristics while suppressing local amplification triggering, ensuring that isolation actions strictly follow the main fault trajectory. This shortens fault response time, reduces the probability of erroneous actions, and improves the timing coordination of fault isolation and the safety and reliability of system operation.

[0044] This invention provides, for example Figure 2 The distribution network fault location system based on multi-source sensor fusion shown includes a high-frequency disturbance recording module, a propagation consistency discrimination module, a high-frequency feature stripping module, a dual-axis fault location module, and an isolation action control module. The high-frequency disturbance recording module collects high-frequency fluctuation signals corresponding to multiple sensor points based on transient high-frequency disturbances generated in the early stage of distribution network faults. It expands each high-frequency fluctuation signal under a unified time base, extracts the start and end times of the protruding segments in the high-frequency fluctuation signals, and forms a locally magnified fingerprint set based on the protruding segments. The propagation consistency discrimination module constructs a propagation consistency constraint relationship based on the local magnified fingerprint set. It binds the arrival order of each protruding segment in the local magnified fingerprint set to the corresponding distribution network topology distance, identifies protruding segments whose arrival order deviates from the distribution network topology distance, and forms a list of suspected false propagation points. The high-frequency feature stripping module, based on the list of suspected pseudo-propagation points, performs bandwidth convergence-type peak clipping on the protruding segments corresponding to the list of suspected pseudo-propagation points, while retaining the energy transition boundary information, forming the main fault trace after removing the local amplification effect. The dual-axis fault location module determines the earliest stable inflection point in the time dimension and maps the shortest consistent segment in the spatial dimension based on the main fault trajectory. It generates a real fault segment lock list based on the earliest stable inflection point and the shortest consistent segment. The isolation action control module, based on the real fault section locking list, introduces an extremely short silent time window within the real fault section to suppress local amplification triggering, and dynamically guides the timing of isolation commands according to the earliest stable inflection point of the main fault trajectory, so that the isolation action unfolds along the main fault trajectory.

[0045] The distribution network fault location method based on multi-source sensor fusion provided in this embodiment of the invention is implemented through the aforementioned distribution network fault location system based on multi-source sensor fusion. For details of the specific methods and processes of the distribution network fault location system based on multi-source sensor fusion, please refer to the embodiments of the distribution network fault location method based on multi-source sensor fusion described above, which will not be repeated here.

[0046] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for fault location in distribution networks based on multi-source sensor fusion, characterized in that, Includes the following steps: To address the transient high-frequency disturbances generated in the early stages of a distribution network fault, high-frequency fluctuation signals corresponding to multiple sensor points are collected. Under a unified time reference, each high-frequency fluctuation signal is expanded, and the start and end times of the spike segments in the high-frequency fluctuation signals are extracted. Based on the spike segments, a locally magnified fingerprint set is formed. Based on the local magnified fingerprint set, a propagation consistency constraint relationship is constructed. The arrival order of each protruding segment in the local magnified fingerprint set is bound to the corresponding distribution network topology distance. Protruding segments whose arrival order deviates from the distribution network topology distance are identified, forming a list of suspected false propagation points. Based on the list of suspected false propagation points, bandwidth convergence peak reduction processing is performed on the protruding segments corresponding to the list of suspected false propagation points, while retaining the energy transition boundary information to form the main fault trace. Based on the main fault trace, the earliest stable inflection point is determined in the time dimension, and the shortest consistent segment is mapped in the spatial dimension. Based on the earliest stable inflection point and the shortest consistent segment, a real fault segment locking list is generated. Based on the real fault section locking list, an extremely short silent time window is introduced within the real fault section to suppress local amplification triggering, and the timing of isolation commands is dynamically guided according to the earliest stable inflection point of the main fault trace.

2. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 1, characterized in that, The steps for forming a locally magnified fingerprint set are as follows: Multiple sensing devices are deployed at the locations of power distribution lines, switching equipment and transformer nodes to collect high-frequency fluctuation signals and to unfold each high-frequency fluctuation signal with a unified time base. The unfolded high-frequency wave signal is continuously scanned in time to extract the spike segments with increased amplitude change rate in the high-frequency wave signal, and the start and end times of the spike segments are determined. The protrusion segments of each sensing point are cross-mapped under a unified time base, and the amplitude change pattern and duration distribution of the protrusion segments are analyzed to form a locally magnified fingerprint unit. The magnified fingerprint units of each sensor point are summarized according to time sequence and spatial location to form a magnified fingerprint set.

3. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 2, characterized in that, The process of forming the list of suspicious points in the false propaganda is as follows: After obtaining the locally magnified fingerprint set, the arrival time of the protruding segments in the locally magnified fingerprint set is mapped to the global time axis under a unified time base, and the spatial location mapping relationship of each sensing point in the distribution network topology is established. An arrival sequence is constructed based on the temporal relationship between multiple protruding segments, and the arrival sequence is bound to the distribution network topology distance level one by one; A correlation analysis was performed on the binding results between the arrival order and the distribution network topology distance to identify abrupt segments where the arrival order and the distribution network topology distance showed a reverse offset. The protruding segments that show reverse offsets are compiled to form a list of suspected false propagation points.

4. The method for fault location in distribution networks based on multi-source sensor fusion according to claim 3, characterized in that, When generating the list of suspected false propagation points, the temporal attributes, topological distance attributes, and locally magnified fingerprint features of the protruding segments are preserved. Under a unified time reference, the temporal continuity and spatial consistency of the protruding segments are maintained, so that the list of suspected false propagation points can accurately reflect the distribution characteristics of non-real propagation segments.

5. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 3, characterized in that, The steps for forming the main fault trace based on the list of suspected false propagation points are as follows: After obtaining the list of suspected false propagation points, the high-frequency fluctuation signals corresponding to the protruding segments in the list of suspected false propagation points are mapped to the time axis under a unified time base, while maintaining the start and end time boundaries of the protruding segments. Based on the mapped protruding segments, bandwidth convergent peak clipping is performed on the protruding segments, and the peak clipping is limited to the start and end time range of the protruding segments. During the bandwidth convergence peak-shaving process, the energy transition boundary information in the spike segments is preserved. The spike segments that have undergone peak shaving and retain energy transition boundary information are integrated with the remaining spike segments according to a unified time reference to form the main fault trace.

6. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 5, characterized in that, During the formation of the main fault trace, the bandwidth convergent peak clipping process uses the energy distribution state of the protruding segment as a constraint to suppress the high-frequency energy peaks inside the protruding segment, while maintaining the continuous existence of the energy transition boundaries corresponding to the start and end positions of the protruding segment. This ensures that the main fault trace remains completely connected in the time dimension and avoids interference from local amplification features in the energy dimension.

7. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 5, characterized in that, The steps for generating a list of actual faulty sections based on the main fault trace are as follows: After the main fault trace is formed, it is unfolded under a unified time base, and the changing trend of the main fault trace is analyzed along the time axis to determine the earliest stable inflection point of continuous change from a stable state to an abnormal state. Based on the earliest stable inflection point, the main fault trace is mapped to the distribution network topology in the spatial dimension, and the distribution of the main fault trace at different spatial locations is unfolded. Based on spatial expansion, the continuous distribution range of the main fault trace in the distribution network topology is compressed and mapped to extract the shortest consistent segment corresponding to the earliest stable inflection point. By combining the earliest stable inflection point with the shortest consistent segment, a list of real faulty segments is generated.

8. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 7, characterized in that, In the process of generating a real fault section lock list, the continuous distribution relationship of the main fault trace in the distribution network topology is used as a constraint condition to ensure that the shortest consistent section maintains a continuous mapping in the spatial dimension and that the earliest stable inflection point is associated with the shortest consistent section in the time dimension.

9. The method for fault location in a distribution network based on multi-source sensor fusion according to claim 7, characterized in that, The steps for performing isolation actions based on the actual faulty section lockout list are as follows: After obtaining the list of real faulty sections, the faulty section information in the list is matched with the time distribution information of the main fault trace, and an extremely short silent time window is introduced at the starting position of the real faulty section. After the extremely short silent time window ends, the timing of isolation commands is dynamically adjusted based on the earliest stable inflection point of the main fault trace. Based on the time rhythm after dynamic traction, the isolation action is carried out sequentially within the actual fault section along the propagation direction of the main fault trajectory. During the isolation process, the isolation actions of each section within the actual fault section are sequentially linked in time to form a sequence of isolation actions that unfolds continuously along the main fault trajectory.

10. A distribution network fault location system based on multi-source sensor fusion, used to implement the distribution network fault location method based on multi-source sensor fusion as described in any one of claims 1-9, characterized in that, It includes a high-frequency disturbance recording module, a propagation consistency discrimination module, a high-frequency feature stripping module, a dual-axis fault location module, and an isolation action control module: The high-frequency disturbance recording module collects high-frequency fluctuation signals corresponding to multiple sensor points based on transient high-frequency disturbances generated in the early stage of distribution network faults. It expands each high-frequency fluctuation signal under a unified time base, extracts the start and end times of the protruding segments in the high-frequency fluctuation signals, and forms a locally magnified fingerprint set based on the protruding segments. The propagation consistency discrimination module constructs a propagation consistency constraint relationship based on the local magnified fingerprint set. It binds the arrival order of each protruding segment in the local magnified fingerprint set to the corresponding distribution network topology distance, identifies protruding segments whose arrival order deviates from the distribution network topology distance, and forms a list of suspected false propagation points. The high-frequency feature stripping module, based on the list of suspected pseudo-propagation points, performs bandwidth convergence-type peak clipping on the protruding segments corresponding to the list of suspected pseudo-propagation points, while retaining the energy transition boundary information to form the main fault trace. The dual-axis fault location module determines the earliest stable inflection point in the time dimension and maps the shortest consistent segment in the spatial dimension based on the main fault trajectory. It generates a real fault segment lock list based on the earliest stable inflection point and the shortest consistent segment. The isolation action control module, based on the real fault section locking list, introduces an extremely short silent time window within the real fault section to suppress local amplification triggering, and dynamically guides the timing of isolation commands according to the earliest stable inflection point of the main fault trace.

Images