Low-voltage power distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion

Abstract

The application discloses a low-voltage distribution network intelligent voltage-regulating energy-saving method based on multi-source data fusion, relates to the technical field of intelligent control of power systems, and comprises the following steps: S001, a transient separation collection band is arranged at an incoming line side, voltage signals and current signals are cut according to millisecond-level time particles, a sharp peak sketch is formed, and a tail of the sharp peak sketch is extracted; S002, the sharp peak sketch is spliced with a smooth running section to form a comparison band, and a resonance fingerprint set is extracted according to the tail of the sharp peak sketch and is used for marking transient resonance characteristics. The application constructs a resonance fingerprint set and a guardrail time axis, avoids misjudgment of transient sharp peaks, and prevents frequent switching of reactive power compensation; through a reactive power buffer bin and a multi-dimensional flexible adjustment means, dynamic compensation is realized, power quality is improved, and energy saving and system stability are taken into account.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for power systems, and specifically to an intelligent voltage regulation and energy-saving method for low-voltage distribution networks based on multi-source data fusion. Background Technology

[0002] Smart voltage regulation and energy saving in low-voltage distribution networks of smart grids based on multi-source data fusion refers to the simultaneous collection of various data from distribution transformers, transformer substations, power terminals, environmental monitoring equipment, and historical operation databases during the operation of low-voltage distribution networks in smart grids. Multi-source data fusion technology integrates heterogeneous information such as voltage, current, load characteristics, line loss indicators, weather factors, and user behavior into a unified state description. Furthermore, intelligent analysis models are used to identify problems such as voltage deviation, load imbalance, and reactive power fluctuations in real time, generating optimal voltage regulation strategies. This allows for coordinated control of tap changer adjustments, reactive power compensation device switching, distributed power generation output, and load-side response. This reduces line losses, minimizes reactive power transmission, and improves energy efficiency while meeting power quality requirements, achieving energy-saving operation and stable regulation of the smart grid distribution network.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, intelligent voltage regulation and reactive power compensation in low-voltage distribution networks typically rely on multi-source data fusion models to identify voltage fluctuations and changes in reactive power demand in real time. However, when a dynamic scenario occurs in the system with the start-up of a high-power impact load, local lines may generate rare resonant spike signals due to abrupt changes in the coupling relationships of inductance, capacitance, and impedance. These signals exhibit high amplitude and short duration characteristics over time. Because existing fusion algorithms lack independent identification and anomaly removal mechanisms for transient resonant characteristics, they often misjudge such resonant spikes as long-term stable trend signals, leading to an abnormal spike in the model's estimation of reactive power demand within a short period. This misjudgment further induces frequent switching of reactive power compensation devices, forming high-frequency current surge cycles. This causes the compensation capacitors to bear overload stress in a short time, resulting in dielectric breakdown, a sharp reduction in lifespan, and even instability of the system compensation circuit, seriously threatening the safety and reliability of the distribution network operation.

[0005] 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

[0006] The purpose of this invention is to provide an intelligent voltage regulation and energy-saving method for low-voltage distribution networks based on multi-source data fusion, so as to solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a smart voltage regulation and energy-saving method for low-voltage distribution networks based on multi-source data fusion, comprising the following steps:

[0008] S001, a transient separation acquisition band is set on the incoming line side, and the voltage signal and current signal are segmented according to millisecond-level time particles to form a peak sketch and extract the tail of the peak sketch;

[0009] S002, the peak sketch is spliced ​​with the smooth running segment to form a contrast band, and the resonance fingerprint set is extracted from the tail of the peak sketch to mark transient resonance characteristics;

[0010] S003, an obstacle avoidance time-series guardrail is established upstream based on the resonant fingerprint set, and a guardrail boundary time axis is constructed along the tail of the resonant fingerprint set to limit the effective range of the transient response;

[0011] S004, a reactive power buffer chamber is constructed around the time axis of the guardrail boundary, and the impact reactive power signal is temporarily stored at the end of the boundary time axis to generate a reactive power buffer trajectory for energy balance adjustment.

[0012] S005 performs reverse breathing window adjustment, shadow compensation adjustment and reversible micro-discharge adjustment based on the reactive power buffer trajectory, and drives the capacitor soft-cut action along the tail of the reactive power buffer trajectory to construct a dynamic control closed loop to maintain steady-state voltage regulation and energy-saving operation of the low-voltage distribution network.

[0013] Preferably, step S001 includes:

[0014] High-frequency acquisition probe groups are deployed on the three-phase conductors on the incoming side of the low-voltage distribution network to acquire voltage and current signals and perform electrical isolation.

[0015] The voltage and current signals are segmented according to the set time granularity, and the maximum value, minimum value, average rising edge slope, and zero-crossing distance of each frame of signal are calculated to form a peak sketch.

[0016] Extract the tail portion of the peak sketch that exhibits natural decay characteristics and encapsulate it into tail segment data packets;

[0017] The tail segment data packet is spliced ​​with the previous smooth running frame to form an abnormal transition segment sketch, which is used for subsequent resonance characteristic analysis.

[0018] Preferably, step S002 includes:

[0019] The data at the tail of the peak sketch is used as the baseline input and stitched together with the data of the continuous smooth running segment on the time axis to generate a contrast band containing the stable segment, the peak segment and the decay segment.

[0020] Waveform analysis is performed on the spliced ​​comparison band to identify oscillation behavior with reverse deflection and decreasing amplitude in the tail signal, and the corresponding oscillation behavior sequence is constructed.

[0021] A resonant fingerprint set is constructed based on the start and end times, maximum amplitude, oscillation duration, number of deflections, and amplitude change rate extracted from the oscillation behavior sequence, and multidimensional feature items are summarized.

[0022] The resonant fingerprint set is compared with the features of historical events, the time coordinates of the transient resonant features are marked, a marker spectrum is formed in the comparison band, and it is used for subsequent control boundary setting.

[0023] Preferably, after splicing the peak sketch with the stable operation section to form a contrast band, the oscillation behavior identification is performed on the three-phase signals respectively, and the triggering phase of the transient event is determined based on the difference in oscillation amplitude of each phase, which is used to indicate the resonance type caused by unbalanced load.

[0024] Preferably, step S003 includes:

[0025] Read the oscillation start time, peak point, duration and three-phase signal offset intensity from the resonant fingerprint set, and construct the avoidance timing window by extending the guardrail time axis forward based on the oscillation end point;

[0026] In the power distribution path corresponding to the guardrail time axis, the three-phase conductor is used as the basic channel, and it extends upstream step by step according to the peak voltage offset, and guardrail marks are superimposed to form a layered guardrail area.

[0027] Freeze all voltage regulation commands within the guardrail time interval, covering tap changer adjustment, reactive power compensation switching and load-side response operations, to prevent control logic from being triggered erroneously in a short period of time.

[0028] After the guardrail time ends, a recovery observation period is set. The freezing is determined based on the voltage and current fluctuation range. The guardrail duration is dynamically adjusted according to the resonant fingerprint expansion rule.

[0029] Preferably, the execution steps of freezing the voltage regulation action command within the guardrail time interval include setting a freezing duration for each upstream node in the guardrail path, and the freezing range gradually decreases with the peak voltage offset.

[0030] Preferably, step S004 includes:

[0031] Call the guardrail boundary time axis information and extract its end time, establish a buffer time window that extends backward, and sample the three-phase reactive power signal in real time within the buffer time window;

[0032] The sampled signals are constructed into a buffered original trajectory in chronological order, and high-risk segments with fluctuation ranges exceeding the preset safety threshold are marked by a sliding window method.

[0033] Waveform data is extracted around the point of maximum disturbance on the buffer time axis to construct a buffer trajectory map containing amplitude peak, oscillation frequency and three-phase response;

[0034] The buffer trajectory map is aligned to the end of the time axis at the guardrail boundary and divided into a disturbance rise segment, a peak control segment, and a decay stabilization segment to guide the segmented response adjustment strategy.

[0035] Preferably, the alignment base point of the buffer trajectory map is the starting point of the first high-risk segment. The adjustment strategy prohibits capacitor switching in the disturbance rising segment, enables the inverse breathing window adjustment in the peak control segment, and performs reversible micro-discharge adjustment in the decay stabilization segment.

[0036] Preferably, step S005 includes:

[0037] Read the reactive power buffer trajectory map and identify the rising segment of the disturbance. Perform reverse breathing window adjustment to briefly absorb excess reactive power signal and limit the rising trend of reactive power.

[0038] Shadow compensation adjustment is implemented within the peak control period. The compensation response curve is planned in advance based on the peak prediction position, and the preset reactive power capacity is gradually released to enhance the lag absorption capacity.

[0039] During the transition from the buffer trajectory to the attenuation section, reversible micro-discharge adjustment is performed, the discharge entry point is set and the release rate is controlled to maintain the continuity of capacitor operation;

[0040] The capacitor soft-cut action is delayed at the end mark of the reactive power buffer trajectory, and the closed-loop control cycle is completed by adopting a step-by-step load reduction method to achieve steady-state voltage regulation and energy-saving operation.

[0041] Preferably, the reactive power absorption operation in the reverse breathing window adjustment needs to be consistent with the phase of the three-phase voltage fluctuation. In the shadow compensation adjustment, there is a response waiting section before the compensation device enters the output range. The capacitor soft-switching action adopts the method of gradually adjusting the access angle to reduce the power factor impact.

[0042] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0043] This invention effectively avoids the problem of traditional fusion models misjudging transient spikes as steady-state trends by constructing a resonant fingerprint set and setting the boundary range of reactive power response based on the guardrail time axis. This prevents high-frequency impacts and capacitor overload risks caused by frequent switching of reactive power compensation devices. Furthermore, by temporarily storing abnormal power through a reactive power buffer and combining flexible control methods such as inverse breathing window adjustment, shadow compensation adjustment, and reversible micro-discharge adjustment, segmented dynamic compensation is achieved during non-steady-state periods. This effectively reduces ineffective energy transmission and equipment losses, thereby improving power quality while achieving energy-saving goals and enhancing the safety, economy, and adaptability of the distribution network. Attached Figure Description

[0044] 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.

[0045] Figure 1 This is a flowchart of the intelligent voltage regulation and energy-saving method for low-voltage distribution networks based on multi-source data fusion, as described in this invention. Detailed Implementation

[0046] 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.

[0047] This invention provides, for example Figure 1 The intelligent voltage regulation and energy-saving method for low-voltage distribution networks based on multi-source data fusion, as shown, includes the following steps:

[0048] S001, a transient separation acquisition band is set on the incoming line side, and the voltage signal and current signal are segmented according to millisecond-level time particles to form a peak sketch and extract the tail of the peak sketch;

[0049] To accurately identify transient voltage and current anomalies caused by the startup of high-power impact loads, a data acquisition structure capable of capturing high-time-resolution changes in electrical quantities needs to be installed on the incoming side of the distribution network to extract the spike characteristics formed by transient resonance. This process includes the following steps:

[0050] High-frequency acquisition probe groups are sequentially deployed on each phase power line of the three-phase conductors on the incoming side of the low-voltage distribution network. The probes synchronously sense voltage and current signals, with a sampling frequency controlled above 20kHz to meet the response requirements for millisecond-level changes. During acquisition, the sampling window length is set to 100 milliseconds for each trigger, with a 20-millisecond overlap between adjacent sampling windows to ensure that complete waveform segments are obtained at the moment of signal spikes. After hardware electrical isolation processing, the acquired signals are converted into digital signal frames within a standard amplitude range to ensure electrical safety and signal integrity in subsequent processing stages.

[0051] The acquired voltage and current signals are divided into frames with a time granularity of 100 milliseconds each. Quantitative calculations are performed on the maximum and minimum values, average rise slope, and zero-crossing intervals within each frame to construct a peak sketch containing peak shape information. The peak sketch is presented in parallel with three-phase data for easy comparison of transient characteristics. If the peak value in a frame increases by more than 80% compared to the previous frame, and the signal duration does not exceed 20 milliseconds, it is considered a possible abnormal peak signal. This criterion is designed in conjunction with typical load change characteristics; for example, when a three-phase asynchronous motor starts, it is often accompanied by an inrush current peak lasting 10-30 milliseconds, which can be effectively captured using this method.

[0052] After identifying spike sketches containing potential transient anomalies, the tail signal is extracted independently. The tail is defined as the naturally decaying segment of the spike shape, typically the 20-millisecond data sequence following the peak point. During this process, the rise and fall rates of the spike shape are combined to exclude short-term amplitude changes caused by power frequency oscillations or voltage dips, thus ensuring that the extracted signal accurately reflects the resonance characteristics. The extracted tail data is encapsulated into tail segment data packets for each of the three phases, exhibiting characteristics such as short-term oscillation, relative aperiodity, and dramatic amplitude changes. These packets become the core raw segments for subsequent resonance identification.

[0053] Based on the tail segment data packets, they are spliced ​​and compared with the previous stable operating frame on the time axis. During splicing, data continuity is maintained and no filtering or smoothing is performed to avoid weakening the true shape of the spikes. The spliced ​​result forms a complete sketch of the abnormal transition segment, used for subsequent construction of the resonance fingerprint database. This sketch retains the reference signal from the previous stable operation and fully reflects the tail waveform after the spike, possessing timing, directionality, and amplitude abrupt change characteristics. By comparing the oscillation frequency distribution and symmetry of the tail segment data in the sketch, the connection location of the transient source, load type, and possible line resonance conditions can be further inferred.

[0054] S002, the peak sketch is spliced ​​with the smooth running segment to form a contrast band, and the resonance fingerprint set is extracted from the tail of the peak sketch to mark transient resonance characteristics;

[0055] To perform in-depth analysis of the spike sketches extracted from voltage and current signals, they need to be spliced ​​with a reference signal under stable operating conditions to construct a contrast band. Based on this, a set of resonant fingerprints with transient characteristics is extracted to accurately label the behavioral features of transient events. This process includes the following steps:

[0056] The acquired tail data of the spike sketch is used as the reference input. This tail data represents a typical waveform of the transient voltage or current response transitioning from high amplitude to steady state, characterized by irregular oscillations, non-periodic symmetry, and amplitude convergence. During the splicing process, a steady-state data segment continuous with the tail time axis of the spike sketch and before any disturbance occurs in the system is selected as the reference segment. The data time length is taken as the first 200 milliseconds to cover at least one full cycle of common low-voltage distribution network voltage fluctuations. To ensure that the signal sequence after splicing is uninterrupted in time, the timestamps of the two data segments must be synchronized, and the voltage level baseline must be calibrated to avoid splicing distortion due to measurement offset. After splicing, a comparison band of no less than 300 milliseconds is generated, which fully covers the entire process from the system's steady state to the occurrence of the resonant spike and then to the tail decay.

[0057] Waveform analysis was performed on the spliced ​​comparison band, analyzing the signal variation patterns sequentially from the tail to the front, identifying the amplitude, period, and phase trends exhibiting oscillation characteristics in the tail section. Specifically, starting with the maximum peak point of the spike sketch, the next 50 milliseconds of data were used as the main analysis segment. The focus was on observing whether the signal fluctuations showed multiple reverse deflections, with the amplitude decreasing relative to the previous deflection after each deflection. If there were three consecutive deflections with a decrease rate exceeding 30%, this signal segment was defined as the resonant core interval. This identification process was performed independently in each of the three phases of data, and the corresponding time segments, phase start values, and number of deflections were marked to construct a corresponding oscillation behavior sequence, laying the foundation for subsequent fingerprint feature extraction.

[0058] After obtaining tail data containing multiple key oscillation behaviors, a resonance fingerprint set is constructed based on the previously extracted results. The construction principle of the resonance fingerprint set is to summarize five categories of indicators in time series form: start and end time, maximum amplitude, oscillation duration, number of deflections, and amplitude change rate. Each fingerprint entry corresponds to a combination of features. For example, if a transient event shows five reverse oscillations with a maximum deviation reaching 22% of the effective voltage value and a duration of approximately 18 milliseconds, this data set is treated as an independent fingerprint record. This fingerprint set not only reflects the temporal and amplitude patterns in electrical characteristics but also possesses certain spatial distribution information. For instance, if one phase shows no significant deviation in the event while the other two phases oscillate strongly, it indicates that the event may have originated from an unbalanced load trigger. By summarizing the multidimensional feature entries of each event, the fingerprint set constitutes a standard reference set that can be used for rapid identification of future transient events.

[0059] In the constructed resonant fingerprint set, representative typical feature groups are selected based on the oscillation amplitude and duration indicators of each record, and compared with known types in the historical event database. If the comparison result shows that there is a historical event record with a matching degree of more than 75%, the fingerprint can be included in the high-confidence fingerprint set to mark the transient resonant features of the current identification event. After marking, the start and end positions of the oscillation are marked in the comparison band in the form of time coordinates to form a marking spectrum. At the same time, color codes are superimposed on the three phase channels to facilitate subsequent visual review and anomaly diagnosis. This marking is not only used for static analysis, but also serves as the boundary basis for the construction of avoidance time sequence guardrails in subsequent steps to ensure that voltage regulation control is not erroneously triggered during short-term abrupt changes.

[0060] S003, an obstacle avoidance time-series guardrail is established upstream based on the resonant fingerprint set, and a guardrail boundary time axis is constructed along the tail of the resonant fingerprint set to limit the effective range of the transient response;

[0061] To prevent erroneous regulation behavior in low-voltage distribution networks during transient resonance, a timing barrier needs to be established upstream of the distribution path based on the extracted resonance fingerprint set. This barrier then defines the time boundaries for response actions to suppress false triggering and frequent operation of reactive power control. This process includes the following steps:

[0062] The extracted resonant fingerprint set is retrieved, and the oscillation start time, peak point, duration, and offset intensity of the three-phase signal corresponding to each fingerprint are read. These fingerprint entries provide the temporal basis and intensity grading criteria for constructing the guardrail. For ease of unified processing, the oscillation end point in each fingerprint record is used as the anchor position, defining it as the tail start point of the guardrail time axis. Then, the leading edge boundary of the guardrail is extended forward by 1.5 times the length of the fingerprint oscillation duration, thereby defining a complete avoidance timing window on the time axis. For example, if the oscillation duration of a certain fingerprint is 20 milliseconds, the corresponding guardrail window length is 30 milliseconds, with the time period starting 10 milliseconds before the peak point and ending 20 milliseconds after the oscillation end point.

[0063] In the power distribution path corresponding to the guardrail timeline, three-phase conductors serve as the basic channel, extending upstream level by level according to the line number corresponding to the identified event. The depth of each upstream extension level can be dynamically adjusted based on the peak voltage offset recorded in the fingerprint set; the larger the offset, the further the extension. For example, if the fingerprint set shows a maximum offset of 22% in a certain event, the guardrail is traced back 3 levels; if the offset is less than 10%, it is traced back only 1 level. This method ensures that the guardrail covers all paths that may be affected by transient interference, rather than creating an indiscriminate protection zone for the entire power distribution network. Guardrail markers are superimposed on each upstream path, recording the corresponding time period and node number to form a multi-layered, directional guardrail zone.

[0064] Based on the constructed upstream path guardrail, voltage regulation behavior within the guardrail section is subject to response restrictions. Specifically, all execution commands related to dynamic voltage and current adjustments, including tap changer adjustments, reactive power compensation switching, and load-side responses, are frozen within the guardrail's time interval. This ensures that control logic is not triggered by transient disturbances within a short period. For example, if a resonant fingerprint window of 40 milliseconds is detected within a certain line segment, the adjustment command signals for that line node and its two upstream nodes are temporarily blocked for 40 milliseconds. The original control state is restored after the window expires. This prevents transient disturbances from misleading the system into identifying trend changes, thereby triggering invalid or reverse adjustment operations.

[0065] During the effective period of the barrier zone, to prevent the backlog of voltage regulation commands or response delays from causing a chain reaction to the system, a barrier buffer release strategy needs to be set. After the barrier zone ends, a recovery observation period of no less than 20 milliseconds is used to resample the voltage and current status of the current line and assess whether it has returned to the normal range. For example, if voltage fluctuations are detected to be no more than ±3% and current fluctuations to be no more than ±5% during the recovery period, the previously frozen regulation actions are allowed to be gradually restored. If high-frequency oscillation signals still exist during the recovery period, the barrier window duration can be appropriately extended according to the fingerprint set expansion rules to avoid hastily restarting the response control before stabilization. The avoidance sequence barrier constructed in this way not only achieves dynamic protection in terms of time, but also, combined with the spatial distribution of the power distribution path and the characteristics of event intensity, forms a precise shield against the transient resonance influence zone, providing stable boundary conditions for subsequent reactive power buffering and flexible control.

[0066] S004, a reactive power buffer chamber is constructed around the time axis of the guardrail boundary, and the impact reactive power signal is temporarily stored at the end of the boundary time axis to generate a reactive power buffer trajectory for energy balance adjustment.

[0067] To handle short-term reactive power fluctuations caused by transient resonance within the protection period provided by the timing barrier, a reactive power buffer chamber needs to be constructed around the time axis of the barrier boundary. This buffer temporarily stores abnormal signals and generates buffer trajectories, providing timing data for subsequent energy balance adjustments. This process includes the following steps:

[0068] The completed guardrail boundary timeline information is retrieved, and the end time of each guardrail segment is extracted as the starting reference point for the buffer chamber. Starting from this time point, a buffer time window extending backwards is established, with an initial length of 80 milliseconds. This length parameter, experimentally determined, can cover the duration of reactive power disturbances caused by impulsive loads in most low-voltage distribution networks. Within this buffer time window, the three-phase reactive power signal is sampled in real time at a sampling frequency of no less than 5 kHz to ensure the capture of peak changes in impulsive reactive power fluctuations. Each sample generates a set of values, corresponding to the inter-phase difference of instantaneous reactive power, the intra-phase fluctuation intensity, and the relative average offset value, totaling three dimensions, forming complete time-series reactive power fluctuation information.

[0069] The collected reactive power signals are normalized within a buffer time window to unify the time and amplitude scales, and arranged into a continuous buffer data series according to the sampling order, forming the original buffer trajectory. To avoid distortion of the overall trajectory by a single spike, a sliding window analysis method is introduced, with segments every 20 milliseconds. The fluctuation range and rate of change are dynamically evaluated. If the fluctuation in a segment exceeds a preset safety threshold (e.g., a sudden increase in the reactive power component exceeding 10 kVar), the segment is marked as a high-risk segment and stored separately for subsequent key adjustments. The number of marked segments in the entire buffer trajectory must be no less than three, and the total time must be no less than 30% of the buffer time window; otherwise, the reactive power fluctuation in that segment is considered insufficient to meet buffering requirements. This judgment ensures that the subsequent energy balancing procedure is only initiated when the reactive power fluctuation actually constitutes adjustment pressure.

[0070] The marked high-risk reactive power segments are extracted on the buffer time axis, with their maximum disturbance point serving as the core point of the trajectory. Waveform data for 20 milliseconds before and after this point are collected to construct a local trajectory feature region. The data in this feature region includes not only the amplitude peak and oscillation frequency, but also the degree of resonance between the three phases at the same frequency and the peak time difference, thus determining the physical cause, such as capacitor intrusion, inductive load startup, or distributed power source inverter interference. Based on these characteristics, a buffer trajectory map is constructed, which shows the complete disturbance development process and three-phase response process, and is marked with time sequence and directionality. The buffer trajectory map will serve as the input reference in the subsequent dynamic adjustment stage, driving the selection of reactive power response strategies and the control of adjustment amplitude.

[0071] After the buffer trajectory is constructed, it is aligned with the end time of the obstacle avoidance sequence barrier. The alignment base point is set as the starting point of the first high-risk segment of the buffer trajectory, ensuring that the adjustment action starts immediately after the barrier opens, avoiding premature intervention or delayed response. The entire buffer trajectory timeline is then divided into three segments: the disturbance rise segment, the peak control segment, and the attenuation stabilization segment. Each segment has a different allowable adjustment range and response priority. For example, capacitor switching is prohibited in the disturbance rise segment, breathing control strategies are allowed in the peak control segment, and fine-tuning compensation strategies are allowed in the attenuation stabilization segment. This segmented adjustment based on the buffer trajectory provides a flexible response time after a transient impact, ensuring the stability of the distribution network is not disrupted while smoothly guiding the system back to steady-state power distribution after the peak. This constructs a reactive power buffer processing mechanism that combines time perception, disturbance identification, and dynamic adjustment, laying the foundation for subsequent closed-loop control.

[0072] S005 performs reverse breathing window adjustment, shadow compensation adjustment and reversible micro-discharge adjustment based on the reactive power buffer trajectory, and drives the capacitor soft-cut action along the tail of the reactive power buffer trajectory to build a dynamic control closed loop to maintain steady-state voltage regulation and energy-saving operation of the low-voltage distribution network.

[0073] To implement refined response control based on the established reactive power buffer trajectory, it is necessary to sequentially execute reverse-phase breathing window adjustment, shadow compensation adjustment, and reversible micro-discharge adjustment based on the characteristics of this trajectory, and drive capacitor soft-switching at the end of the buffer to form a closed-loop dynamic control mechanism, thereby maintaining the steady-state voltage regulation of the distribution network and reducing energy loss. This process includes the following steps:

[0074] Read the generated reactive power buffer trajectory diagram, paying particular attention to the three marked trajectory segments: the disturbance rise segment, the peak control segment, and the attenuation and stabilization segment. At the beginning of the disturbance rise segment, initiate the reverse-phase breathing window adjustment operation. This operation briefly absorbs excess reactive power signal by setting a virtual power window in the opposite direction, enabling the system to suppress the impact without changing the main loop topology. For example, if the rise segment lasts for 40 milliseconds and the reactive power rises to 1.8 times the normal value, then within this window, artificially inject power with a reverse absorption capacity of 20% to offset it, forming a negative feedback effect and limiting further reactive power increase. The entire reverse absorption process must be kept in phase with the three-phase voltage fluctuation to prevent additional current imbalance caused by phase misalignment.

[0075] After entering the peak control phase, a shadow compensation adjustment method is implemented. This method presets a virtual compensation behavior path within the compensation channel outside the main circuit load path. Instead of immediately outputting the physical compensation amount, it pre-plans the response curve based on the predicted peak position in the reactive power buffer trajectory. Taking a certain impact event as an example, if the reactive power peak appears at the 65th millisecond of the buffer time axis, with an amplitude of 2.4 times the normal state, a compensation response command is preset at the 55th millisecond, keeping the compensation equipment in standby mode. It begins to enter the output stability range 5 milliseconds before the peak, gradually releasing the pre-set reactive power capacity. This ensures that the response process has hysteresis absorption capability, avoiding system voltage fluctuations caused by response delay. Shadow compensation adjustment not only improves response accuracy but also reserves margin for subsequent discharge operations.

[0076] During the transition from the peak to the attenuation phase, a reversible micro-discharge regulation strategy is initiated. This strategy sets a discharge entry point at the tail of each high-risk segment in the buffer trajectory and sets a gradually decreasing discharge rate, allowing the capacitor element to slowly release excess stored energy without stopping operation. For example, if the tail-end fluctuation amplitude is still 10% higher than the average level, the discharge path is opened, initially releasing 8% of the total capacity every 10 milliseconds, gradually slowing down to 2% when the fluctuation amplitude is below 5%. This micro-discharge method avoids the impact loss caused by sudden capacitor element shutdown and also avoids the need for re-switching due to rapid reactive power decline, thereby extending the service life of the compensation equipment while ensuring power supply stability. The entire discharge process must be coordinated with the voltage fluctuation trend and cannot be initiated independently to ensure synchronization and coordination in the regulation closed loop.

[0077] At the end marker of the reactive power buffer trajectory, a soft-switching action is initiated to complete the closed loop of the entire dynamic regulation. The soft-switching action must be performed only after the compensation capacity has stabilized and the discharge phase is complete. Its initiation timing is chosen to be 20 milliseconds after the buffer trajectory ends, serving as an observation and confirmation period. The soft-switching process employs a step-by-step load reduction method, for example, removing one capacitor unit every 5 milliseconds, and gradually adjusting the connection angle to reduce its impact on the system power factor until all actions are completed and the system returns to its pre-regulation steady-state operating level. During this period, voltage and current fluctuation indicators are monitored to ensure that all indicators return to within the allowable error range before the entire dynamic regulation cycle is considered complete. This method ensures a smooth transition during the regulation process while effectively avoiding problems such as surge impacts and capacitive reactance jumps that occur during traditional capacitor switching, providing strong support for the long-term stable operation of the distribution network and laying the foundation for achieving energy-saving goals.

[0078] This invention effectively avoids the problem of traditional fusion models misjudging transient spikes as steady-state trends by constructing a resonant fingerprint set and setting the boundary range of reactive power response based on the guardrail time axis. This prevents high-frequency impacts and capacitor overload risks caused by frequent switching of reactive power compensation devices. Furthermore, by temporarily storing abnormal power through a reactive power buffer and combining flexible control methods such as inverse breathing window adjustment, shadow compensation adjustment, and reversible micro-discharge adjustment, segmented dynamic compensation is achieved during non-steady-state periods. This effectively reduces ineffective energy transmission and equipment losses, thereby improving power quality while achieving energy-saving goals and enhancing the safety, economy, and adaptability of the distribution network.

[0079] 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 low-voltage power distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion, characterized in that, The method comprises the following steps: S001, transient separation acquisition band is set at the incoming line side, voltage signals and current signals are divided according to millisecond level time particles, a peak sketch is formed, and a tail of the peak sketch is extracted; S002, the peak sketch is spliced with a smooth running section to form a comparison band, a resonance fingerprint set is extracted according to the tail of the peak sketch, and is used for marking a transient resonance feature; S003, an avoidance time fence is established based on the resonance fingerprint set at the upstream, and a fence boundary time axis is constructed along the tail of the resonance fingerprint set, which is used for limiting an action interval of a transient response; S004, a reactive power buffer bin is constructed around the fence boundary time axis, an impact reactive power signal is temporarily stored at the tail of the boundary time axis, a reactive power buffer track is generated to perform energy balance adjustment; S005, based on the reactive power buffer track, inverse breathing window adjustment, shadow compensation adjustment and reversible micro-discharge adjustment are performed, and a capacitor soft cut action is driven along the tail of the reactive power buffer track, a dynamic regulation closed loop is constructed, and steady state voltage regulation and energy saving operation of a low voltage distribution network are maintained; Step S005 comprises: reading a reactive power buffer track atlas and identifying a disturbance rising section, performing inverse breathing window adjustment to temporarily absorb excess reactive power signals, and limiting the rising trend of reactive power; in the peak control section, shadow compensation adjustment is implemented, a compensation response curve is planned in advance according to the peak prediction position, the preset reactive power capacity is gradually released, and the lag absorption capacity is enhanced; in the transition process of the buffer track to the decay section, reversible micro-discharge adjustment is performed, the discharge entry point is set and the release rate is controlled to maintain the continuity of capacitor operation; at the end mark of the tail of the reactive power buffer track, the capacitor soft cut action is started with delay, the closed loop regulation cycle is completed in a step-by-step load shedding manner, and steady state voltage regulation and energy saving operation are realized. 2.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 1, characterized in that, Step S001 comprises: high-frequency acquisition probes are arranged on the three-phase conductors at the incoming line side of the low voltage distribution network, voltage signals and current signals are collected and electrically isolated; the voltage signals and current signals are divided according to the set time granularity, the maximum value, minimum value, average slope of the rising edge and the distance between zero-crossing points of each frame of signal are calculated to form a peak sketch; the tail of the peak sketch which appears natural decay characteristics is extracted and encapsulated as a tail section data packet; the tail section data packet is spliced with the previous stable running frame to form an abnormal transition section sketch, which is used for subsequent resonance feature analysis. 3.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 2, characterized in that, Step S002 comprises: the tail section data of the peak sketch is input as a reference, and is spliced with the stable running section data of the time axis to generate a comparison band containing a stable section, a peak section and a decay section; the comparison band after splicing is analyzed, the oscillation behavior with reverse deflection and amplitude decrease in the tail section signal is identified, and the corresponding oscillation behavior sequence is constructed; a resonance fingerprint set is constructed according to the start and end time, maximum amplitude, oscillation duration, deflection number and amplitude change rate extracted from the oscillation behavior sequence, and multi-dimensional feature items are summarized; the resonance fingerprint set is compared with historical event features, the time coordinates of the transient resonance feature are marked, a mark atlas is formed in the comparison band, and is used for subsequent control boundary setting.

4. The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion according to claim 3, characterized in that, The spike sketch is spliced with the stable running section to form a contrast band, and oscillation behavior identification is performed on the three-phase signals, and the triggering phase of the transient event is determined based on the oscillation amplitude difference of each phase, which is used to indicate the resonance type caused by unbalanced load.

5. The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion according to claim 3, characterized in that, Step S003 comprises: Read the oscillation start time, peak point, duration and three-phase signal offset strength in the resonance fingerprint set, and extend the guardrail time axis based on the oscillation end point to construct an avoidance timing window; In the distribution path corresponding to the guardrail time axis, take the three-phase conductor as the basic channel, extend upstream according to the peak voltage offset, and superimpose the guardrail mark to form a layered guardrail area; Freeze all voltage regulating action commands in the guardrail time interval, cover the tap changer adjustment, reactive power compensation switching and load side response operation, and prevent false triggering of control logic in a short time; After the end of the guardrail time, set a recovery observation section, determine whether to unfreeze according to the voltage and current fluctuation range, and dynamically adjust the guardrail time according to the resonance fingerprint extension rule. 6.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 5, characterized in that, The execution step of freezing voltage regulating action commands in the guardrail time interval includes setting a freezing time for each upstream node in the guardrail path, and the freezing range decreases step by step with the peak voltage offset amplitude. 7.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 5, characterized in that, Step S004 comprises: Call the guardrail boundary time axis information and extract its end time, establish a backward extending buffer time window, and sample the three-phase reactive power signals in real time in the buffer time window; Construct a buffer original trajectory according to the time sequence of the sampling signals, and mark the high-risk fragments whose fluctuation range exceeds the preset safety threshold through sliding window method; Extract waveform data around the maximum disturbance point on the buffer time axis to construct a buffer trajectory atlas containing amplitude peak, oscillation frequency and three-phase response; Align the buffer trajectory atlas to the end time of the guardrail boundary time axis, and divide it into disturbance rising section, peak control section and decay stable section, which are used to guide the segmented response adjustment strategy. 8.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 7, characterized in that, The alignment base point of the buffer trajectory atlas is the start point of the first high-risk fragment, the adjustment strategy prohibits capacitor switching in the disturbance rising section, enables anti-phase breathing window adjustment in the peak control section, and executes reversible micro-discharge adjustment in the decay stable section. 9.The low-voltage distribution network intelligent voltage regulation and energy saving method based on multi-source data fusion of claim 1, characterized in that, The reactive power absorption operation in the anti-phase breathing window adjustment needs to be consistent with the phase of three-phase voltage fluctuation, and there is a response waiting section before the compensation device enters the output interval in the shadow compensation adjustment, and the capacitor soft switching action uses the step-by-step adjustment of the access angle to reduce the power factor impact.

Images