A method and system for regulating three-phase imbalance of low-voltage cable branch boxes

Abstract

The application discloses a three-phase imbalance adjustment method and system for a low-voltage cable branch box, which identifies a structural imbalance section and judges a heavy load phase by monitoring three-phase current and phase-to-phase voltage; the dispatching state of a neighboring branch box of a same feeder is inquired before commutation to avoid conflict, and the fault identification margin of a superior protection device after commutation is verified, and accordingly, an executable commutation object is determined; after commutation, actual migration effect is verified according to the power consumption characteristics change of a downstream user electric energy meter; and the tightening and loosening standards of the determination threshold are dynamically adjusted according to the recurrence situation within a threshold relaxation period after commutation. The method distinguishes between load driving and structural imbalance, cooperatively avoids dispatching conflicts between devices, and respectively verifies the protection margin before and after commutation and the load migration effect, thereby reducing invalid dispatching caused by misjudgment or commutation conflict, and reducing the situation that needs to be repeatedly disposed due to protection cooperation failure or load not being truly transferred after commutation.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution technology, and in particular to a method and system for adjusting three-phase imbalance in a low-voltage cable branch box. Background Technology

[0002] In low-voltage power distribution networks, a large number of single-phase loads are distributed and connected to three-phase lines. In actual power consumption, it is difficult to maintain the balance of three-phase loads. Long-term three-phase imbalance will lead to an increase in neutral current, an increase in line loss, and a deviation of some phase voltages from the rated value. It will also aggravate the local heating of transformers and switching equipment, affecting power supply quality and equipment lifespan.

[0003] Existing adjustment methods mostly rely on manual adjustments to wiring after periodic inspections and load statistics. This results in a slow response time and difficulty in adapting to real-time load changes. While some branch box equipment with automatic phase switching capabilities can determine whether adjustments are needed based on instantaneous current data, they typically do not consider whether there are scheduling conflicts between adjacent equipment on the same feeder, whether the fault identification capability of the upper-level protection device remains reliable after phase switching, or the means to verify the actual effect of phase switching. This can easily lead to problems such as repeated scheduling of adjacent equipment, failure of protection coordination, or the load not being truly transferred after adjustment. Summary of the Invention

[0004] This invention discloses a method and system for adjusting three-phase imbalance in low-voltage cable branch boxes. The aim is to collaboratively identify the three-phase load imbalance in the branch box, implement phase switching adjustment while avoiding scheduling conflicts between adjacent equipment and ensuring reliable operation of the protection device after phase switching, and verify the phase switching effect, thereby improving the adaptability and reliability of three-phase imbalance management.

[0005] A first aspect of the present invention discloses a method for adjusting three-phase imbalance in a low-voltage cable branch box, comprising: Operating parameters are obtained by collecting three-phase current and phase-to-phase voltage through the in-box current transformer. Based on the operating parameters, the structural imbalance section is determined by analyzing the no-load segment deviation and zero-load baseline drift. For the structural imbalance section, analyze the characteristics of the heavy-load branch to generate a heavy-load phase identifier. Based on the heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes on the same feeder to obtain the target phase conflict identifier. According to the target phase conflict identifier, select non-conflicting light-load phases to form a candidate switch identifier group. The over-limit duration ratio is calculated by the candidate switch identification group to obtain the over-limit day proximity. The over-limit day proximity is graded according to the urgency to obtain the screening tolerance threshold. The target switch identification is obtained by evaluating the operation margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group. Based on the target switch identifier, the action parameters are determined to form switching control parameters. The switching control parameters are transmitted to the on / off state of the switch output contacts. Based on the on / off state of the contacts, the correlation of the electricity consumption characteristics of the user-side electricity meter is analyzed to obtain the load migration identifier. Based on the number of recurrences that still exceed the limit after the load migration identifier statistical threshold is relaxed, an asymmetric adjustment coefficient is generated. Based on the asymmetric adjustment coefficient, a differentiated periodic parameter is converted to generate a dynamic judgment threshold. Based on the dynamic judgment threshold, the load status is judged and a load adjustment command is output.

[0006] A second aspect of the present invention discloses a three-phase imbalance adjustment system for a low-voltage cable branch box, comprising: The status identification unit is used to acquire three-phase current and phase-to-phase voltage through the in-box current transformer to obtain operating parameters, and to determine the structural imbalance section based on the operating parameters by performing no-load segment deviation analysis and zero-load baseline drift analysis. The identifier generation unit is used to analyze the characteristics of the heavy-load branch in the structural imbalance section to generate a heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes of the same feeder based on the heavy-load phase identifier to obtain the target phase conflict identifier, and select non-conflicting light-load phases to form a candidate switch identifier group according to the target phase conflict identifier. The target selection unit is used to calculate the over-limit duration ratio through the candidate switch identification group to obtain the over-limit day proximity, and to calculate the screening tolerance threshold according to the urgency level based on the over-limit day proximity, and to evaluate the operation margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group to obtain the target switch identification. The switching execution unit is used to determine the action parameters based on the target switch identifier to form switching control parameters, transmit the switching control parameters to the on / off state of the switch output contact, and analyze the correlation of the user-side electricity meter's electricity consumption characteristics based on the on / off state of the contact to obtain a load migration identifier. The instruction output unit is used to generate an asymmetric adjustment coefficient based on the number of recurrences that still exceed the limit after the load migration identifier statistical threshold is relaxed, to generate a dynamic judgment threshold by converting the differentiated cycle parameter based on the asymmetric adjustment coefficient, and to output a load adjustment instruction based on the load status judgment based on the dynamic judgment threshold.

[0007] The beneficial effects of this invention are reflected in the following points: First, this invention does not rely solely on the load current at a certain moment to determine whether there is a three-phase imbalance. Instead, it uses the period when the load approaches zero to check the residual deviation of the three-phase current separately, distinguishing whether the imbalance originates from structural problems of the line itself or is caused solely by current load fluctuations. Based on this, before determining the phase that needs adjustment, it also proactively checks whether there are any unexecuted scheduling plans in adjacent branch boxes on the same feeder, avoiding multiple devices making independent decisions and resulting in superimposed loads or mutual conflicts on the same phase. Second, phase switching may seem like just switching a switch, but the fault current characteristics of that phase will also change after the switch. Before selecting a specific phase switching switch, this invention calculates whether the margin of the upper-level protection device in distinguishing between fault current and normal load current is still sufficient after the phase switching, and adjusts the stringency of this margin requirement according to the urgency of the current imbalance. When urgent, it appropriately relaxes the margin to deal with the problem as soon as possible, and strictly controls the margin when it is not urgent, avoiding blind spots in the originally reliable fault protection due to phase switching. Finally, the completion of the phase switching command does not mean that the load has been truly transferred. This invention verifies whether the corresponding load has indeed migrated by comparing the changes in the electricity consumption characteristics of downstream users' electricity meters before and after the phase switching, in order to discover situations where the ledger records do not match the actual wiring on site. For switches that repeatedly exceed the limit after phase switching, this invention does not use a uniform and fixed judgment threshold, but rather gradually lowers the threshold for tightening monitoring with the number of recurrences and gradually raises the threshold for relaxing monitoring, so that the system maintains higher vigilance for equipment that repeatedly has problems, while not over-intervening in equipment that performs stably. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a three-phase imbalance adjustment method for a low-voltage cable branch box according to the present invention.

[0009] Figure 2 This is a schematic diagram of the scheduling conflict identification and concurrent candidate determination of adjacent branch boxes of the feeder according to the present invention.

[0010] Figure 3 This is a schematic diagram of the dynamic relaxation of the screening tolerance threshold driven by the proximity of the over-limit day in this invention.

[0011] Figure 4 This is a structural block diagram of a three-phase imbalance adjustment system for a low-voltage cable branch box according to the present invention.

[0012] Wherein: 1-Branch box; 2-Feeder; 3-Phase terminal; 4-Communication link; 5-Concurrent phase selection; 6-Conflict judgment box; 7-Execution time window; 8-Total load bar; 9-Proximity to over-limit day axis; 10-Filtering tolerance threshold axis; 11-Constant flat segment; 12-Linear descent segment; 13-Bottoming lower limit segment; 14-Value example point. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0015] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0016] The technical solutions of the embodiments of this application will be described below.

[0017] like Figure 1 As shown, this embodiment of the invention provides a method for adjusting the three-phase imbalance of a low-voltage cable branch box, including the following steps S101-S105: Step S101: Obtain operating parameters by collecting three-phase current and phase-to-phase voltage through the in-box transformer, and determine the structural imbalance section by performing no-load segment deviation analysis and zero-load baseline drift analysis based on the operating parameters.

[0018] Specifically, operating parameters are obtained by collecting three-phase current and phase-to-phase voltage through in-box current transformers. Three-phase current is collected using three current transformers connected in series in the A, B, and C phase outgoing circuits, respectively. The transformation ratio is selected based on the maximum downstream capacity of the branch box, with an accuracy class of 0.5. The secondary side output is connected to the current sampling channel of the monitoring device. Phase-to-phase voltage is collected using voltage transformers in the box, simultaneously acquiring the three line voltage values ​​Uab, Ubc, and Uca, with an accuracy class of 0.2. The voltage transformers and current transformers share the same clock reference, and the current and voltage at each sampling point correspond to the same moment. The clock reference is maintained by a crystal oscillator built into the monitoring device, with a daily cumulative error not exceeding 1 second. Each operating parameter record includes the three-phase currents Ia, Ib, and Ic and the three line voltage values ​​at the same moment. The sampling period is once per minute, and the record includes a timestamp of the acquisition time, with timestamp accuracy to the second. If an abnormal signal from a current transformer in a certain phase prevents the current in that phase from being read, the current in that phase will be recorded as null and marked as abnormal. The other two phases and voltage values ​​will still be written normally, and the entire record will not be discarded. If the transformer ratio does not match the factory calibration value and the nameplate marking, the current value calculated from the operating parameters will have a systematic deviation. This deviation will carry over to all subsequent stages without being easily detected. The three current transformers and voltage transformers will be kept synchronously under the same sampling pulse, and the three-phase current and three line voltages will belong to the same moment. The single-phase sampling deviation will not exceed half a sampling cycle. Records with a delay of more than 200 milliseconds will be additionally marked with a delay label. Their current and voltage values ​​will still be written normally without being affected. Operating parameters marked with a delay label will not be included in the continuous time period during the no-load segment screening to prevent insufficiently timely operating parameters from being mixed into the baseline.

[0019] In some embodiments, the step of determining structural imbalance sections by analyzing the no-load segment deviation and zero-load baseline drift based on the operating parameters includes: screening periods when the load current is lower than the no-load threshold from the operating parameters to obtain no-load segment data; statistically analyzing the three-phase current deviation fluctuations based on the no-load segment data to obtain no-load deviation fluctuation values; generating a baseline drift amount by comparing the no-load deviation fluctuation value with a preset zero-load baseline interval; and determining structural imbalance sections based on the baseline drift amount exceeding the allowable range of sensing errors.

[0020] The process involves screening operating parameters for periods where the load current is below the no-load threshold to obtain no-load segment data. For each operating parameter record, the maximum value of its three-phase current is compared with the no-load threshold. Records with a maximum value lower than the no-load threshold are considered candidates for no-load periods. The no-load threshold is the upper limit of the current corresponding to the typical standby load downstream of the branch box of this model, calculated as 3% of the downstream access capacity. Candidate records must also last for at least 10 minutes. Low-current segments shorter than 10 minutes are more likely to be due to short-term downtime of downstream equipment and are not included in the candidates. Within 10 minutes, a single sampling point is allowed to briefly exceed the threshold. As long as the conditions are continuously met before and after this point, the timing of the candidate period will not be interrupted, and the point exceeding the threshold is still recorded. Only operating parameter records that meet all conditions are retained for no-load segment data. The raw data is recorded and stored, and processing and conversion are left to subsequent stages. Each record is also marked with the start and end times of its candidate period. This stage does not preset a fixed time window and scans operating parameters around the clock. True no-load events occurring outside of late night are also included, as are no-load events caused by downstream users being away during the day on holidays. If the cumulative duration of no-load segment data is less than 30 minutes on a given day, it will not be included in subsequent statistics for that day. A minimum sample size of 30 sampling points is required for deviation fluctuation calculation, based on a 1-minute sampling cycle of 30 minutes. The cumulative duration is obtained by adding the durations of all candidate time periods on that day, and does not need to be concentrated in the same continuous time period. For the first 3 days after a new branch box is put into operation, no no-load segment data will be generated by default, allowing downstream users' electricity usage habits to stabilize. After 3 days, screening will begin according to normal rules. During these 3 days, operating parameters will still be collected and written at a 1-minute cycle as usual, but will not participate in this stage of no-load screening. This avoids including unstable electricity usage patterns during the new installation period in the baseline, and mistakenly treating fluctuations during the new installation adaptation period as zero-load residual deviation.

[0021] The no-load deviation fluctuation value is obtained by statistically analyzing the three-phase current deviation fluctuations based on the no-load segment data. For each sampling point in the no-load segment data, the difference between the maximum and minimum values ​​of the three-phase current is taken as the single-point deviation for that sampling point, in amperes. Theoretically, the three-phase currents should all approach zero under no-load conditions, and there should be no significant difference between them. Therefore, the single-point deviation is the direct basis for judging whether the no-load deviation fluctuation value is abnormal. The no-load deviation fluctuation value is calculated using the formula ΔI=(Σd_k) / n, where d_k is the single-point deviation of the k-th sampling point in the no-load segment data, n is the total number of sampling points in the no-load segment data for that day, and ΔI is the no-load deviation fluctuation value, in amperes, which is the arithmetic mean of the single-point deviations of all no-load sampling points for that day. If an isolated value exists in the no-load segment data where the single-point deviation exceeds three times the median single-point deviation of the day, it will be treated as eddy current noise or transient interference within the box and removed when calculating ΔI. The number of removed values ​​should not exceed 5% of the total number of sampling points on that day. If the number exceeds 5%, all values ​​will be retained for calculation and will no longer be considered noise. The no-load deviation fluctuation value is calculated once a day and is not merged across days. The downstream access equipment may have changed on different days, and merging across days will include equipment changes as structural deviations. If the number of sampling points in the no-load segment data on a certain day is exactly equal to the minimum sample size requirement, ΔI will still be calculated normally according to the formula, without additional weighting or correction due to the sample size just barely meeting the limit. After the no-load deviation fluctuation value is calculated, it will be registered as an independent value according to the branch box number and date. The ΔI values ​​of different days for the same branch box are independent and do not overlap with each other. The ΔI values ​​between different branch boxes are also not referenced to each other because the downstream access equipment and secondary circuit conditions of each branch box are different. If the number of remaining sampling points after removing isolated values ​​is less than the minimum sample size, ΔI is marked as insufficient and excluded from the statistics for that day according to the aforementioned rules. This day is not included in the comparison range of subsequent baseline drift.

[0022] Baseline drift is generated based on the comparison of no-load deviation fluctuation values ​​with a preset zero-load baseline range. The daily no-load deviation fluctuation value is compared with the zero-load baseline range [0, I_ref]. I_ref is the theoretical maximum residual deviation determined by the accuracy class of the current transformer and the secondary circuit contact resistance, calculated by multiplying the rated secondary current of the current transformer by its accuracy class and then applying voltage division based on the contact resistance. The lower limit of the range is fixed at 0. When the no-load deviation fluctuation value is lower than or equal to I_ref, it is considered to fall within the baseline range, and no drift is generated, recorded as 0. When the no-load deviation fluctuation value is higher than I_ref, the baseline drift is calculated using the formula ΔD = ΔI - I_ref, where ΔI is the daily no-load deviation fluctuation value, I_ref is the upper limit of the baseline range for the branch box, and ΔD is the baseline drift, with the same unit as ΔI. A larger value indicates a greater deviation of the no-load current from the theoretical residual deviation. The baseline drift is calculated independently each day. When ΔI does not exceed I_ref on a given day, ΔD is recorded as 0, leaving no blank space. Once the accuracy class of the current transformer and the secondary circuit wiring are determined, the I_ref of the same branch box is fixed. The same reference is used for all dates during comparison to ensure comparability of baseline drift amounts across different dates. When ΔI is higher than I_ref, the baseline drift is linearly calculated based on the excess portion; the greater the excess, the larger the drift. In the critical case where ΔI equals I_ref, the drift is 0, not negative. Negative deviations are not included in the drift amount. Since the lower limit of the baseline interval is fixed at 0, deviations below the baseline are considered to fall within the interval. Therefore, the baseline drift amount obtained only reflects the positive excess and serves as the sole input for determining whether the limit has been exceeded in the next stage.

[0023] For example, determining a structural imbalance segment based on the baseline drift exceeding the allowable range of sensing error includes: generating an initial threshold indication based on the baseline drift with reference to a preset allowable error range; counting the number of repeated threshold violations in consecutive cycles based on the initial threshold indication to obtain a continuous threshold violation count; confirming that a preset continuous threshold has been reached based on the continuous threshold violation count to form a drift confirmation indicator; and excluding occasional false alarm events to determine the structural imbalance segment based on the drift confirmation indicator.

[0024] A preliminary over-limit indicator is generated based on the baseline drift amount, referencing a preset error tolerance range. The error tolerance range is 20% of I_ref. A baseline drift exceeding 20% ​​of I_ref on a given day is considered a preliminary over-limit characteristic. 20% corresponds to the upper limit of the combined assembly tolerance of the same batch of transformers and the secondary circuit temperature drift. Deviations below this percentage are attributed to inherent device errors and are not considered abnormal. The preliminary over-limit indicator is generated when the baseline drift exceeds the error tolerance range on a given day. The indicator records the trigger date and the corresponding ΔD value, and this record is written to the monitoring device's local storage. No indicator or placeholder record is generated on days when ΔD does not exceed the error tolerance range. If the baseline drift of the same branch box exceeds the error tolerance range for several consecutive days, a separate preliminary over-limit indicator is generated for each day, and these are not merged into a single record covering multiple days. Each record allows for independent verification of the ΔD value for the corresponding date. The initial limit violation flag only reflects the limit violation situation on a single day. Whether a structural imbalance is confirmed depends on the situation over subsequent consecutive days. The single-day flag does not trigger any on-site handling actions. If the ΔD is calculated twice on the same day due to data retransmission, only the initial limit violation flag generated the first time will be retained, and the second calculation result will not generate a new flag. In addition to recording the trigger date and the corresponding ΔD value, the initial limit violation flag also calculates the daily average of the residual current of each phase in the no-load segment data for that day. The phase with the highest average value is taken as the main concentrated phase of deviation for that day, and the average residual current and the number of occurrences for that day are recorded. This uses daily accumulation rather than single-point results to avoid the interference of instantaneous noise on the phase. The phase with the higher the daily average value is, the more likely it is to be judged as the long-term concentrated phase after accumulation over multiple days.

[0025] The continuous over-limit count is obtained by counting the number of consecutive over-limit violations based on the initial over-limit judgment flag. A daily check is performed to see if an initial over-limit judgment flag is generated. If generated, the count is incremented by 1; otherwise, it is reset to zero and re-accumulated. After resetting to zero, the next time an initial over-limit judgment flag is generated, the count restarts from 1. The increment or decrement of the count is determined solely by whether a flag is generated on that day, without considering the value of ΔD itself. The continuous over-limit count only counts the number of consecutive days. A single day without an over-limit violation is considered an interruption, and the accumulated days before the interruption are not traced. The count on the day of the interruption is reset to zero, not halved or decreased. If a flag is regenerated the day after the interruption, the count restarts from 1. The continuous over-limit count is updated daily, with the update time matching the generation time of the initial over-limit judgment flag for that day, without lag to the next day. If no flag is generated on a given day, the update action is to set the count to 0. If a branch box's continuous over-limit count reaches a high value and is reset to zero on a particular day without an over-limit violation, the re-counting does not refer to the historical peak value before resetting. Any subsequent consecutive over-limit violations after resetting to zero are treated as a completely new count. The continuous over-limit count itself has no upper limit. Even if it far exceeds the threshold required for subsequent confirmation, it continues to accumulate until a day when it does not exceed the limit is reached, at which point it is reset to zero. During the accumulation process, the value of each day is bound to the date and archived. Resetting the count to zero and keeping the count unchanged are two independent actions within the system. The former writes the value to 0, while the latter does not rewrite the value. Both are triggered by the presence or absence of a daily marker. If there is a marker on the current day, the value is incremented by 1 based on the previous day's value. If there is no marker on the current day, it is directly written to 0. This avoids confusion between the two situations of "no marker" and "missing data" in subsequent processes, as the two situations have different meanings for determining the drift confirmation marker in the next process.

[0026] A drift confirmation flag is generated based on the continuous over-limit count reaching a preset continuous threshold. The preset continuous threshold is 7 days. Confirmation is triggered when the continuous over-limit count reaches 7. 7 days covers a complete natural week's combination of weekdays and rest days, avoiding occasional cyclical fluctuations that only occur on weekdays or weekends. If the period is shorter than 7 days, unilateral fluctuations on rest days are easily misconfirmed. The drift confirmation flag is generated on the day the continuous over-limit count first reaches 7. The flag contains the date of the trigger and the average ΔD value of the most recent 7 days. The average value is obtained by adding the ΔD values ​​of each of the 7 days and dividing by 7. After the continuous over-limit count continues to exceed 7, a new drift confirmation flag is not generated every day. It is only generated once when the count changes from below 7 to 7. From the 8th day onwards, the continuous over-limit count is only updated, and no second flag is generated. Once a drift confirmation flag is generated, it is retained in the record. Even if the continuous over-limit count is reset to zero the next day due to interruption, the generated drift confirmation flag is not revoked. If the same branch box reaches the 7-day threshold twice in different months, an independent drift confirmation flag is generated for each. The two records correspond to the average ΔD value of the most recent 7 days at the time of each trigger. In addition to the date and the average ΔD value, the drift confirmation flag also summarizes the main phase of the daily deviation in the preliminary judgment flag records of the 7-day limit exceedance. If the concentrated phase always points to the same phase within 7 days, the phase number is included in the flag as a preliminary basis for the long-term existence of abnormality in that phase. If the concentrated phase is not fixed within 7 days, this item is recorded as uncertain, indicating that the no-load residual deviation of the branch box does not stably point to a single phase, but alternates between two or three phases. This alternation pattern usually corresponds to the presence of slight contact problems in multiple terminals of the secondary circuit, or multiple current transformers in the same batch being at the edge of the accuracy band, which is different in cause from the situation of a single phase having a long-term high deviation.

[0027] The drift confirmation flag is used to exclude occasional false alarms and determine structural imbalance sections. After the drift confirmation flag is generated, it is first checked whether there are any commutation operations or transformer replacement records for the branch box within the past 7 days. The verification is based on the work order records retained by the field operation system. If such records exist, the drift confirmation flag is judged as a transient false alarm caused by operational disturbance and is not included in the structural imbalance section. If the same branch box has been removed from the structural imbalance section before and the drift confirmation flag is triggered again this time, it is considered to be included again and still needs to go through the same work order verification process. The work order records used for verification are taken from the operation entries of the branch box in the field operation system for the past 7 days, including both commutation and transformer replacement operation times. As long as either operation time falls within the 7-day window, the current confirmation is judged as an operational disturbance, and its ΔD average value is not written into the structural imbalance section. The work order verification not only excludes transient deviations caused by changes in contact state after commutation, but also excludes false residual currents caused by open circuits in the secondary circuit during transformer replacement. Each record in the structural imbalance section also carries the concentrated phase number summarized in the corresponding drift confirmation identifier. Branch boxes with uncertain concentrated phases will still generate records as usual, and this field will be left blank. For branch boxes that have been verified to have no related operations, their numbers and concentrated phases will be written into the structural imbalance section as input objects for analyzing the characteristics of overloaded branches. The average ΔD value of the most recent 7 days will be archived with the records for operation and maintenance to check the degree of imbalance, but will not participate in subsequent phase selection. Records determined to be operational disturbances will be registered with the corresponding work order number and excluded from the current round of sections. Their numbers will not enter the subsequent phase switching process in this round. They will be included again after the drift confirmation identifier is triggered again and verified. When included, the average ΔD value of the most recent 7 days will be recalculated according to the trigger time of the new round, and the value when it was excluded will not be used.

[0028] Step S102: Analyze the characteristics of the heavy-load branch in the structural imbalance section to generate a heavy-load phase identifier. Based on the heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes of the same feeder to obtain the target phase conflict identifier. Select the non-conflicting light-load phase to form a candidate switch identifier group according to the target phase conflict identifier.

[0029] Specifically, heavy-load phase identifiers are generated by analyzing the characteristics of heavily loaded branches in structurally unbalanced sections. For example... Figure 2As shown, for each structural imbalance section, the corresponding branch box 1 is recorded. The three-phase currents for all non-no-load periods (i.e., operating parameter records where the maximum three-phase current is not lower than the no-load threshold) over the past 7 days are taken. The arithmetic mean is calculated phase by phase to obtain the average load current of each of the three phases of branch box 1. Each phase corresponds to a phase terminal 3 of branch box 1. The phase proportion is calculated using the formula P_x = I_x / (Ia + Ib + Ic), where x is one of the three phases A, B, and C corresponding to phase terminal 3, I_x is the average load current of that phase over the past 7 days, and P_x is the proportion of that phase in the total three-phase load. Theoretically, P_x should be close to 33.3% when the three phases are balanced. When the highest proportion of the three phases exceeds 40%, that phase is considered a heavily loaded phase. 40% corresponds to a single phase deviating by 20% from the balanced value, and the neutral current reaching the level of 30% of the phase current. Deviations within 7 percentage points of the theoretical balanced value are not considered heavily loaded. The heavy-load phase identifier records the branch box number and the identified heavy-load phase number, along with the average load current I_x and P_x values ​​for each of the three phases, and the start and end dates of the 7-day interval used for calculation. I_x for each phase is used to calculate the superimposed load after the target phase shift, and P_x for each phase is used to compare the proportions among the light-load phases. For branch boxes where the highest proportion of any of the three phases is less than 40%, no heavy-load phase identifier is generated in this round, even if the branch box is listed in a structural imbalance section. The existence of structural problems reflected in a structural imbalance section and whether the current load is concentrated in a single phase are two independent matters. If a branch box has less than 24 hours of non-no-load time accumulated in the past 7 days, the sample size is insufficient to support stable phase proportion estimation, and a heavy-load phase identifier will not be determined in this round. The phase proportion is calculated based on the cumulative current of the past 7 days rather than a single-point instantaneous value, which can mitigate short-term proportion jumps caused by intermittent start-stop of single-phase equipment. 24 hours corresponds to the minimum data amount for at least one day of full-load operation within 7 days.

[0030] In some embodiments, the step of identifying the scheduling conflict status of adjacent branch boxes on the same feeder based on the heavy-load phase identifier to obtain the target phase conflict identifier includes: querying the communication address of adjacent branch boxes on the same feeder based on the heavy-load phase identifier to obtain the neighbor box communication link; reading the phase switching instruction to be executed based on the neighbor box communication link to generate the neighbor box suspension intention; identifying candidate switching target phases concurrently based on the neighbor box suspension intention to form concurrent conflict identifiers; and verifying preset conflict conditions based on the concurrent conflict identifiers to obtain the target phase conflict identifier.

[0031] The communication links between adjacent branch boxes are obtained by querying the communication addresses of adjacent branch boxes on the same feeder for the heavy-load phase identifier. Branch box 1 corresponding to the heavy-load phase identifier first queries the locally preset feeder 2 topology table. The topology table records the feeder number to which each branch box belongs and the communication addresses of each branch box within the feeder. Branch boxes under the same feeder 2 are numbered according to their electrical position on the feeder. Adjacent branch boxes are listed in adjacent addresses in the table, allowing direct reading of adjacent branch boxes on both sides of this box without traversing the entire table. The table is then searched by branch box 1 number in the heavy-load phase identifier to retrieve the communication addresses of all branch boxes 1 under the same feeder 2 number, excluding this box. The search results are arranged in ascending order of address. The search itself is completed within the monitoring device and does not require additional network requests. Each adjacent branch box communication link 4 corresponds to one adjacent branch box 1. The link establishment uses the carrier communication module built into this type of branch box 1, with a fixed communication rate of 9600 baud, which is not adjusted due to changes in the number of branch boxes 1 on feeder 2. The communication module shares the same power supply as the aforementioned current transformers in the monitoring device and does not require independent power. It is extremely rare, but not uncommon, for the number of adjacent branch boxes within the same feeder to be zero. In this case, the communication link between adjacent boxes is empty, and the conflict identification process is skipped, treating it as a conflict-free situation. The commutation operation corresponding to this overloaded phase identifier is not subject to the constraints of adjacent box coordination. Once established, the adjacent box communication link remains connected and is not disconnected due to a single query. Carrier communication is transmitted on the low-voltage cable of the same feeder, eliminating the need for additional communication lines. At a baud rate of 9600, reading the commutation command of one adjacent branch box on a single link takes approximately several hundred milliseconds. Multiple adjacent branch boxes within the feeder are read sequentially according to their addresses, without concurrent access, to avoid carrier channel conflicts.

[0032] The process generates a neighboring box suspension intention based on reading the phase-switching command to be executed from the neighboring box communication link. Queries are initiated sequentially along each neighboring box communication link to read whether a phase-switching command has been generated but not yet issued for execution in the corresponding adjacent branch box. This command, read back via the neighboring box communication link, represents an intermediate result after the adjacent branch box itself has run this method. This step only reads and does not modify the command; the reading action does not trigger recalculation or premature execution. For adjacent branch boxes with a phase-switching command to be executed, the command content includes the rated capacity of the downstream branch to be switched, the target phase, and the scheduled execution time. These three items are included in the query results via the neighboring box communication link. All three items are taken from the original command text already generated locally in the adjacent branch box; no secondary conversion is performed in this step. The neighboring box suspension intention includes the target phase value, downstream branch rated capacity, and scheduled execution time for each adjacent branch box with a phase-switching command to be executed in the query results, arranged by adjacent branch box number, without deduplication or merging. Even if two adjacent branch boxes happen to select the same target phase, they are recorded separately. If a communication link between neighboring boxes times out and fails to receive a response during the query process, the neighboring branch box is recorded as an "unknown" state in this query. This does not imply that it has no pending phase switching instructions. In concurrent conflict determination, an "unknown" state is considered equivalent to the branch box selecting any light-load phase other than the heavy-load phase of its own box, and is included in the conflict statistics to avoid underestimating conflicts due to missed readings. If a neighboring branch box itself has no pending phase switching instructions, it is not included in the neighboring box suspension intention count. The number of entries varies depending on the actual number of neighboring branch boxes with suspension intentions. If all neighboring branch boxes have no suspension intentions in this query, the neighboring box suspension intention list is empty. An empty list itself is also a valid query result and is not considered a query failure.

[0033] Based on the neighboring box suspension intention, candidate switching target phases are identified and concurrent conflict flags are generated. For each of the three phases (A, B, and C), the number of adjacent branch boxes with that phase as the target phase in the neighboring box suspension intention is counted. This count scans each entry in the neighboring box suspension intention phase by phase. When the same target phase is selected by two or more adjacent branch boxes simultaneously, the scheduled execution times of the phase switching instructions of these branch boxes are further compared. If the scheduled execution times fall within the same 30-minute window, it is determined as a concurrent candidate. The determination of concurrent candidates is performed separately for each target phase, without affecting the determination results of other target phases. If fewer than two adjacent branch boxes are selected for a target phase, it is not determined as a concurrent candidate, and no corresponding concurrent conflict flag is generated; the target phase is treated as having no concurrency. If only the target phases are repeated (i.e., concurrent phase selection 5 is the same), but the scheduled execution times differ by more than 30 minutes from the execution time window 7, they are not considered concurrent candidates. Staggered execution itself avoids short-term concentrated impacts. The 30-minute execution time window 7 is taken from the upper limit of the time required for the voltage of the same feeder segment to recover to a steady state after one commutation disturbance. A commutation time difference of more than 30 minutes between two units ensures that the previous disturbance subsides before the next execution. This value comes from the equipment's factory performance data. The concurrent conflict identifier marks the target phase number that meets the concurrent candidate condition, along with a list of corresponding adjacent branch box numbers. One concurrent conflict identifier corresponds to one target phase. Different target phases are marked independently. The same adjacent branch box may appear simultaneously in two different target phase lists. The marked content is not deduplicated. Each list simultaneously records the scheduled execution time and the rated capacity of the downstream branch for each adjacent branch box.

[0034] The target phase conflict identifier is obtained by verifying the preset conflict conditions for concurrent conflict identifiers. The preset conflict conditions are that the number of adjacent branch boxes involved in the concurrent conflict identifier of the target phase is three or more, or the total downstream load current of the involved branch boxes exceeds 80% of the rated current of the feeder section of the target phase. Meeting either condition is considered a real conflict. The two conditions are verified independently; meeting either one is sufficient to determine a conflict. Three boxes indicate that three loads are simultaneously switched on, exceeding the controllable range of a single dispatch. 80% corresponds to the usable upper limit of the feeder section's rated current after deducting short-term fluctuation margin. The quantity condition corresponds to the range circled in conflict judgment box 6, and the load condition corresponds to the percentage shown in total load bar 8. Concurrent situations involving only two branch boxes with a small total load are not considered conflicts, as the design margin of this feeder section is determined during the feeder design phase and will not be adjusted temporarily in this stage. The target phase conflict identifier only takes effect when the concurrent conflict identifier meets the preset conflict conditions, and the corresponding target phase number is marked. When selecting and switching target phases in this box, the phase marked as conflict must be avoided. The marking content also retains the number of adjacent branch boxes and the total load current value at the time of triggering. The total load current is calculated by summing the rated capacity currents of the downstream branches of each adjacent branch box in the concurrent conflict identification list, and comparing it with the rated current of the feeder section using the same caliber. Both are calculated based on the effective value of the line current. If the concurrent conflict identification does not meet the preset conflict conditions, this step marks the target phase as available in the set of available target phases in this box and records its current total load current; phases that generate target phase conflict identifications after meeting the conflict conditions are marked as restricted in this set. The set of available target phases is arranged from low to high total load current. The conflicting phase has been marked as restricted in the set, forming a complete criterion for whether the two lightly loaded phases in this box can be selected, and the load weight when selected.

[0035] Based on the target phase conflict identifier, candidate switch identifier groups are formed by prioritizing non-conflicting lightly loaded phases. Among the two phases other than the heavy-load phase corresponding to the heavy-load phase identifier of this branch box, phases marked as conflicting by the target phase conflict identifier are first excluded. If neither phase is marked as conflicting, the phase with the lower phase proportion P_x is selected as the target phase. If the proportions of the two phases are equal, the phase with the later phase sequence is selected first. If only one phase remains unmarked as conflicting, that phase is directly selected as the target phase. If both phases are marked as conflicting by the target phase conflict identifier, no candidate switch identifier group is generated in this round. This branch box has no selectable phase switching path and must wait until the phase switching command of the adjacent branch box is executed and the target phase conflict identifier is cleared before making a final determination. The candidate switch identification group includes all switch numbers downstream of the branch box that are currently connected to the heavy-load phase corresponding to the heavy-load phase identifier and whose rated capacity does not exceed the remaining capacity of the target phase. The remaining capacity of the target phase is the difference between the rated current capacity of the feeder section of that phase and the capacity corresponding to the equivalent upstream load current. Switches with rated capacities exceeding this difference are not included to avoid the target phase becoming a new heavy-load phase after switching. If all switch numbers under the heavy-load phase of the same branch box fail to be included in the candidate switch identification group due to capacity reasons, the branch box is recorded as having no feasible switches in this round, and a switch that would cause the target phase to overload will not be forcibly included. Each switch in the candidate switch identification group is labeled with its current phase, the target phase to be switched to, and the downstream rated capacity. The current phase is used to verify that the switch is indeed connected to the heavy-load phase identified in this case. The target phase to be switched to is used to calculate the superimposed load of the target phase after the transfer. The downstream rated capacity is used to calculate the remaining capacity of the target phase and the switching action parameters. All three items are written into the log for each switch in this step. If any item is missing in the log, it should be filled in first before being written in. Incomplete entries are not used as placeholders.

[0036] Step S103: Calculate the over-limit duration ratio through the candidate switch identification group to obtain the over-limit day proximity. Based on the over-limit day proximity, calculate the screening tolerance threshold according to the urgency level. Based on the screening tolerance threshold and the candidate switch identification group, evaluate the operation margin of the upstream protection device of each candidate switch to obtain the target switch identification.

[0037] Specifically, the proximity of the over-limit day is obtained by calculating the over-limit duration percentage through the candidate switch identification group. For each switch number in the candidate switch identification group, the current sampling record of the downstream branch of that switch for the most recent day is taken. The current of this branch is synchronously collected by the branch current transformer connected to the outgoing side of the corresponding switch at a 1-minute cycle with the same operating parameters. The load current of each sampling point is compared with the rated current converted from the rated capacity of the downstream of the switch. Sampling points that exceed the rated current are counted as over-limit sampling points. The over-limit duration percentage is calculated according to the formula R_d=t_over / 1440, where t_over is the cumulative number of minutes corresponding to the over-limit sampling points on the day, 1440 is the total number of minutes in a day, and R_d is the over-limit duration percentage for the day, with a value ranging from 0 to 1. R_d is 0 when there are no over-limit sampling points throughout the day, and R_d is 1 when there are continuous over-limits throughout the day. R_d is calculated and archived daily according to the natural day, and each candidate switch obtains an R_d value for each day. The proximity of the over-limit day is calculated using the formula W=ΣR_d, where W is the sum of R_d for the most recent 7 days. A larger cumulative value indicates that the over-limit period of the switch is closer to a full day. The theoretical maximum value is 7, corresponding to 7 consecutive days of full-day over-limit. The proximity of the over-limit day is calculated for each switch individually. The W values ​​of multiple switches in the same branch box's candidate switch group do not affect each other, even if they are connected to the same heavily loaded phase. This is because the power consumption periods of downstream loads differ for different switches. For switches with missing data within the most recent 7 days, the W calculation skips those days' R_d values ​​and only sums the days with available data. Missing days are not included in the denominator and are not padded with zeros. The resulting W is then converted to a 7-day caliber based on the number of valid days before comparison. Switches with less than 2 valid days have their over-limit day proximity calculated as invalid in this round and are not included in subsequent screening tolerance threshold calculations.

[0038] The screening tolerance threshold is calculated based on the urgency level of the proximity to the limit date. When the proximity W exceeds 2.0, for every 1.0 increase, the screening tolerance threshold decreases by 5 percentage points along the linear decreasing segment 12. This decrease relaxes the qualification standard, and the threshold drops to a minimum of 10% corresponding to the bottom limit segment 13 before ceasing further reduction. 10% is the minimum allowable margin for this type of protection device in emergency scenarios, and this lower limit is given by the manufacturer's technical manual. The benchmark screening tolerance threshold is 20% (corresponding to the constant level segment 11), meaning that the subsequent fault identification margin of the candidate switch must be no less than 20% to be considered qualified. This benchmark corresponds to the minimum margin required for the normal operation of this type of protection device and is the default starting point before this adjustment. When the proximity of the overdue date does not exceed 2.0, the screening tolerance threshold remains unchanged at the baseline value of 20% (i.e., constant level 11). This range corresponds to situations where there is no urgent need for handling. 2.0 is the turning point between constant level 11 and linearly decreasing level 12, corresponding to approximately two full-day equivalent overdue periods accumulated in the last 7 days. Exceeding this level is considered that the overdue period of the switch has become persistent, and the handling pace needs to be accelerated. The screening tolerance threshold is calculated separately for each switch based on its own proximity of the overdue date. Multiple candidate switches under the same branch box will receive different qualification thresholds due to different proximity of the overdue date. Switches with higher proximity of the overdue date will receive more lenient screening standards, while switches with lower proximity will still be calculated according to the baseline or close to the baseline standards. Figure 3 The horizontal axis represents the proximity of the overdue date (axis 9), and the vertical axis represents the screening tolerance threshold (axis 10). Example point 14 shows the proximity of the overdue date for a certain switch and the corresponding screening tolerance threshold, illustrating the specific application of this method to differentiate screening criteria for switches with different levels of urgency.

[0039] In some embodiments, the step of evaluating the operating margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group to obtain the target switch identification includes: verifying the protection setting position of each candidate switch according to the candidate switch identification group to obtain the protection setting value; calculating the fault current margin after transfer based on the protection setting value to determine the fault identification margin; detecting that the fault identification margin exceeds the reliable operation limit by comparing it with the screening tolerance threshold to generate a protection risk identification; and excluding candidate switches with fault identification abnormalities according to the protection risk identification to obtain the target switch identification.

[0040] The protection setting value is obtained by verifying the protection setting positions of each candidate switch based on the candidate switch identification group. For each candidate switch in the candidate switch identification group, the current setting position configuration of its upstream protection device (usually the incoming circuit breaker in the branch box where the switch is located or the fuse in the corresponding circuit in the upstream distribution box) is queried. The configuration record is stored in the ledger of the protection device locally or in the upstream distribution box. Each candidate switch corresponds to the nearest upstream protection device. The protection setting value is taken as the actual operating current value corresponding to the queried setting position, in amperes. Different setting positions of the same model of protection device correspond to different operating currents. It is necessary to verify each device according to the actual setting position. The factory default setting of the model is not used as a substitute, because the actual setting position on site may have been adjusted according to the connected load. The operating current of circuit breaker protection devices is taken as its long-delay trip setting current. The operating current of fuse protection devices is taken as the agreed operating current corresponding to its rated current and fusing characteristics. The two types of devices have different diameters, but both are converted to the effective value of line current for comparison. If the query results are inconsistent with the ledger records, the readings of the setting device measured on-site shall prevail, and the protection setting values ​​shall be corrected accordingly. The operating current corresponding to the setting position shall be converted according to the fixed correspondence between the setting position and the operating current of that type of protection device. The operating current of the same setting is consistent on different devices. Multiple candidate switches in the same branch box may share the same upstream protection device. In this case, the protection setting values ​​of each switch are taken from the reading of the same device, but the superimposed loads after the transfer are different. Therefore, the margin still needs to be calculated for each switch separately. If the ledger record of the upstream protection device of a candidate switch is missing and cannot be verified on-site in a timely manner, the protection setting value of that switch in this round shall be recorded as pending and shall not be used in subsequent calculations without any assumed values. During the pending period, the switch shall not be included in the list of valid candidates in this stage.

[0041] The fault identification margin is determined based on the fault current margin calculated after the transfer according to the protection setting value. The protection setting value serves as a benchmark for assessing risk after the transfer. It is first compared with the expected load current after the transfer. The expected load current is estimated as the sum of the target equivalent load current and the rated load current downstream of this switch. That is, it assumes the result of the load being superimposed on the existing load of the target phase after the switch is switched to the target phase. The target equivalent load current is taken from the I_x value used when calculating the phase proportion in the most recent evaluation cycle of this branch box. The fault identification margin is calculated using the formula M=(I_set-L) / I_set×100%, where I_set is the operating current corresponding to the protection setting value, L is the expected load current after the transfer, and M is the fault identification margin in percentage form. A larger M indicates that the protection device can more easily distinguish the actual fault current from the normal load current after the transfer. When calculating L, if the most recent data for the target equivalent load current is missing, the most recent valid reading for that phase is used instead, and the margin calculation of the entire switch is not skipped due to missing data. The expected load current is calculated by directly adding the existing load current of the target phase to the rated current downstream of this switch, without taking into account the phase difference after the two phases are combined. This is a conservative estimate, and the actual superimposed current is generally not higher than this arithmetic sum. Therefore, the M calculated in this way is too small and the judgment is too strict, and it will not overestimate the margin after the transfer. The fault identification margin is calculated for each switch separately. If there are multiple candidate switches to be switched under the same target phase, they are calculated one by one according to the superimposed result after their respective switching. It is not assumed that only one switch will actually be switched, because the actual dispatch may approve multiple switches to switch to the same target phase at the same time. Theoretically, M can be negative, that is, the expected load current after the transfer has exceeded the protection setting value. In this case, the fault identification margin is recorded as negative without being truncated to 0. This negative value will make the fault identification margin lower than the screening tolerance threshold. Based on this, this step judges the switch as risk and excludes it, and it will not be included in the margin ranking.

[0042] A protection risk flag is generated based on the detection of fault identification margin exceeding the screening tolerance threshold. The fault identification margin M of each switch is compared with the corresponding screening tolerance threshold. If M is lower than the screening tolerance threshold, the protection device's fault identification capability is deemed risky after the switch is transferred, and a protection risk flag is generated. The judgment uses a strict less than rather than less than or equal to threshold; if M is exactly equal to the screening tolerance threshold, the flag is not triggered to avoid the jitter of repeated triggering and cancellation at the boundary value. The reliable operation upper limit is not a fixed value, but a relative criterion that fluctuates with the screening tolerance threshold of each switch. The same M value may lead to different risk judgments for switches with different urgency levels; the more urgent the switch, the more lenient the criterion, and the less urgent the switch, the more stringent the criterion. This floating mechanism is the very meaning of the screening tolerance threshold itself. A large fault identification margin M value indicates that the protection device has sufficient margin to distinguish between fault current and normal load current. M below the threshold means that the normal load after transfer is approaching the set operating current. Once a short-circuit fault occurs, the fault current is superimposed on the already increased load current, and the protection device has difficulty distinguishing between the two near the operating current, which may result in failure to operate or false operation, thus being judged as risky. The switch number that triggered the risk identification record, the M value at that time, and the corresponding screening tolerance threshold are all written into the judgment result of this round. Switches whose M value reaches or exceeds the screening tolerance threshold are considered to be able to reliably identify faults after the protection device is transferred. In this step, their numbers and M values ​​are included in the qualified switch set. The switches in the qualified switch set are sorted from high to low according to their M values. The switch at the top is the one with the highest M value and the most sufficient protection margin after the transfer. Based on this, this step directly selects the top switch in the set as the target switch.

[0043] The target switch identifier is obtained by eliminating fault-identifying candidate switches based on protection risk indicators. Switches with protection risk indicators in the candidate switch identifier group are eliminated in this round and do not participate in the determination of the target switch. The remaining switches without risk indicators proceed to the next comparison. The elimination action triggered by the protection risk indicator only affects the determination in this round and does not change the original record of the switch in the candidate switch identifier group. Among the remaining switches, the one with the highest fault identification margin M is selected as the target switch identifier. The highest M indicates that the fault identification capability of the corresponding protection device of the switch is the most sufficient after the transfer. The switch is selected first for phase switching because it retains the largest protection margin after the switch and has the lowest risk of subsequent false tripping or failure to operate. When the M values ​​are tied for the highest, the switch with the smaller rated capacity is selected first. The smaller capacity switch has relatively fewer downstream connected users, the impact of the switch on downstream users is relatively smaller, and the load increase caused by its transfer to the target phase is smaller. If all switches in the candidate switch identifier group are marked with protection risk indicators and there are no remaining switches to choose from, no target switch identifier is generated in this round, and the dispatch of this branch box is terminated. This situation indicates that any switch transfer under the heavy load phase will result in insufficient protection margin for the target phase. The root cause is often that the target phase itself has insufficient capacity, requiring capacity recalculation or protection setting adjustments before it can be included in the phase switching process. The target switch identification record includes the selected switch number, target phase number, corresponding M value, and the branch box number. The target phase number is taken from the candidate switch identification group and is assigned to the intended target phase for switching, and is transmitted along with the switch number to the switching control parameter stage. If the target phase changes due to concurrent conflict identification limitations, since M depends on the target equivalent preceding load current, the M value for the same switch must be recalculated based on the existing load of the new target phase.

[0044] Step S104: Based on the target switch identifier, determine the action parameters to form switching control parameters, transmit the switching control parameters to the on / off state of the switch output contacts, and analyze the correlation of the user-side electricity meter's electricity consumption characteristics based on the on / off state of the contacts to obtain the load migration identifier.

[0045] Specifically, the switching control parameters are formed by determining the action parameters based on the target switch identification. The physical switch corresponding to the switch number recorded in the target switch identification is a double-throw phase-change switch. The action parameters first determine the target action time, taking the period with the lowest historical average load current of the same hour for the past 30 days as the action time, which usually falls between 2:00 AM and 4:00 AM. The lower the load, the smaller the impact of switching on the power experience of downstream users. The action parameters are calculated using the formula T_gap=200+50×N_load, where 200 is in milliseconds, 50 is in milliseconds per unit, N_load is the estimated number of downstream connected electrical devices of the switch, which is the ratio of the downstream rated capacity marked by the switch in the candidate switch identification group to the rated capacity of a typical single-phase device downstream of this type of branch box, rounded down. T_gap is the interval between the original phase disconnection and the target phase closure, in milliseconds. The more devices there are, the longer the arc extinguishing and contact debouncing time is required, and the longer the interval is accordingly. The switching control parameters include four items: the target action time, the open-close interval T_gap, the switch number in the target switch identifier, and the target phase number. The parameter setup is considered complete only after all four items are written; if any item is missing, the parameter will not be issued. If the target action time has expired (e.g., it is past 4 AM when the calculation is completed), it will be postponed to the same time period the following day, without changing the target switch identifier itself. After the switching control parameters are established, they enter the pending issuance queue, which is ordered according to the target action time. During this time period, the downstream load tends to be at standby level, the voltage drop caused by a single switch is minimal, and the impact on user power consumption is negligible. Multiple switches in the same branch box in the queue will operate sequentially within this time period to avoid the instantaneous superposition of multiple switches in the same branch box.

[0046] The switching control parameters are transmitted to the on / off state of the switch output contacts. After the switching control parameters reach the target action time, the monitoring device sends a disconnect command to the original phase contact via the local control bus. The control bus uses the same carrier communication module as the adjacent box communication link. After the contact performs the disconnect action, it sends back an execution confirmation signal. Upon receiving the confirmation signal, the monitoring device starts the T_gap timer. After the T_gap timer ends, a closing command is sent to the target phase contact (the T_gap value is derived from the switching control parameters). After the contact performs the closing action, it also sends back an execution confirmation signal. Both confirmation signals are timestamped, with timestamp accuracy to milliseconds. The on / off state of the contacts records the switch number corresponding to this switch, the original phase disconnect time, the target phase closing time, and the actual interval between the two times. The actual interval may have a delay of several milliseconds compared to the T_gap setting value. The double-throw phase-changing switch always maintains a gap between disconnection and closure; at any given time, at most one phase is connected, physically preventing the original phase and the target phase from being connected in parallel via the switch. The actual interval is the measured duration of this gap. If no confirmation signal is received within 500 milliseconds after a disconnect command is issued, the switchover execution is deemed abnormal. The contact open / close status is marked as abnormal, and subsequent closing command issuance is suspended. The abnormality information is simultaneously reported to the maintenance system. The original phase contacts remain in the open state, and power restoration requires on-site confirmation and manual handling by maintenance personnel. If no confirmation signal is received within 500 milliseconds after a closing command is issued, the switchover execution is also deemed abnormal. In this case, the original phase is open, but the target phase has not been confirmed closed, and the user is in a power-off state. The contact open / close status is marked as a closing abnormality, and a closing command is immediately reissued. If the reissue also fails, it is prioritized as an emergency alarm and will not wait for the next evaluation cycle.

[0047] In some embodiments, the step of analyzing the correlation of user-side electricity consumption characteristics based on the contact on / off state to obtain a load migration identifier includes: parsing the corresponding downstream branch user account number list from the contact on / off state to establish a target user list; retrieving the power curves of the electricity meter before and after the switch based on the target user list to form a user power curve; verifying that the sudden change at the switch time is in line with expectations based on the user power curve to obtain a power change identifier; and evaluating the branch load correlation strength based on the power change identifier to obtain a load migration identifier.

[0048] A target user list is established by analyzing the downstream branch user account numbers based on the contact on / off status. For switches marked as "normal" (not abnormally terminated) in the contact on / off status record, the corresponding target switch number is retrieved. The user access ledger for the downstream branch of that switch is then queried. The ledger records the downstream user account number and the electricity meter number for each user corresponding to each switch number. One switch number corresponds to all users connected in series on its downstream branch. Users and electricity meters are registered one by one in the access order. The target user list includes all queried user account numbers and corresponding electricity meter numbers, arranged in order of account number. The number of downstream users for the same switch varies depending on the actual number of users connected to that branch; there is no fixed number of entries, typically ranging from single digits to a dozen or so. If a user has cancelled their account but the record has not yet been cleared, that account number will still appear in the query results, and the target user list will still include it. For users whose electricity meters have been cancelled, the lack of remote meter reading data will prevent them from being identified and excluded during subsequent power curve retrieval. This eliminates the need for prior screening at this stage, preventing the omission of actual users due to outdated records. For switches marked as abnormally terminated, no corresponding target user list is generated at this stage. Since the abnormal switch itself has not yet completed phase transition, the user's actual electricity consumption characteristics will not change, making it impossible to verify any correlation. When the same switch triggers a re-issuance of commands due to a closure anomaly within a short period, resulting in two adjacent contact on / off status records, the target user list is only established based on the record that was ultimately successfully executed. The failed attempt before the re-issuance is not used to generate a separate target user list, preventing the same batch of users from being included in two evaluations.

[0049] Based on the target user list, the power curves of the electricity meters before and after the switchover are retrieved to form the user power curve. For each electricity meter number in the target user list, minute-level active power readings are retrieved for one hour before and after the switchover time, totaling 120 data points. The data source is the remote meter reading interface of the user-side smart electricity meter, which actively reports a reading every minute. The user power curve is formed by stitching together the 60 data points before and after the switchover in chronological order. The switchover time itself corresponds to the 60th and 61st data points on the curve. The time base for both data segments is based on the local clock of the electricity meter. If a communication delay occurs near the switchover time, resulting in fewer than 120 data points being obtained, the user power curve is constructed based on the actual obtained data points, and the gap is marked as blank in the curve. If the blank falls exactly between the 60th and 61st points corresponding to the switchover time, the power change at the moment of switchover cannot be determined for that user, it is marked as missing instantaneous data and transferred to the next round of review, and is not included in the current power change flag determination. User power curves are retrieved and archived independently for each household. Curves from different users under the same target switch are not merged. Even if switching occurs at the same time, each curve is recorded as an independent curve. Each curve is accompanied by the corresponding electricity meter number and the target switch number in the target user list. If no corresponding remote meter reading record is found for an electricity meter number in the target user list, that household is not included in the denominator and numerator of the current correlation strength calculation. It is only registered separately in the load migration identifier as data unavailable, ensuring that the correlation strength ratio is calculated only within the range of users with meter reading data. This avoids households without data being mistakenly counted as unresponsive, thus lowering the ratio.

[0050] Power mutation identifiers are obtained based on the user power curve to verify that the sudden change at the switching moment meets expectations. Between points 60 and 61 on the user power curve corresponding to the switching moment, if the power reading shows a significant single-point drop followed by a rebound at the sampling point, and the drop point falls within the same minute as the switching moment, it is determined to be a power mutation that meets expectations. This drop represents the dip left on the minute-level meter reading due to the second-level power outage experienced by the user at the moment of switching. Since the switching itself takes much less time than a sampling period, only the presence of this dip, rather than its duration, is used as the criterion. The power mutation identifier accurately marks each household with one of three states: expected, partially related, or no response, along with the corresponding user number and the specific time of the mutation. If the power reading on the user power curve remains consistently stable before and after the switching moment, without any pattern of dropping to near zero, it is determined that the user's actual electricity consumption is likely not on the branch in this switching, and it is marked as no response. This situation itself is a signal that the ledger record may be inconsistent with the actual wiring. If the power reading does show a brief drop but the drop is not close to zero (e.g., only half the normal level), it is determined that some load on the user's branch is still in the original phase. This is marked as partially related and pending verification, and is distinguished from the unresponsive status using different status codes in the record. Power surge indicators are determined independently for each household, with conclusions drawn solely based on the user's own power curve performance. Although all households on the same branch experience switching at the same time, whether there was electricity usage at the moment of switching and whether the meter reading happened to cover that minute vary. Therefore, the status of each household must be interpreted separately. The absence of a dip in the status of an individual household does not negate the success of the switching for the entire branch.

[0051] The load migration identifier is obtained by assessing the load association strength of the branch based on the power mutation identifier. The ratio of the number of users in the target user list that generate the expected power mutation identifier to the total number of users in the list is the load association strength of the branch switched by the target switch. The closer the ratio is to 1, the more reliable the electrical association between the downstream users recorded in the ledger and the actual switched branch. The ratio itself ranges from 0 to 100%. The load migration identifier records the target switch number, the association strength ratio, and the verification results of each household (one of three states: expected, partial association pending verification, no response) taken from the power mutation identifier of each household. The percentage of households in each of the three states is recorded together for subsequent differentiation between individual user abnormalities and overall branch ledger inaccuracies. The subsequent handling methods for the two situations are different - individual user abnormalities only require verification of the individual household wiring, while overall branch inaccuracies require re-verification of the ledger of all users downstream of the switch. For load migration indicators below 70%, a ledger verification suggestion will be added, indicating to maintenance personnel that there may be errors in the user access ledger for this branch, requiring on-site verification. The 70% figure is taken from the data accuracy acceptance requirements during the construction of this type of ledger system; values ​​below the acceptance requirements are considered to indicate that the ledger quality itself is substandard. For load migration indicators reaching or exceeding 70%, no ledger verification suggestion will be generated; only routine verification results will be recorded. Users marked as partially associated in the load migration indicator will still be recorded separately, even if the overall association strength meets the standard. Partial association is manifested as a partial decrease in power at the switching point rather than a complete drop, indicating that part of the user's load migrates with this switch while another part remains in the original phase. This corresponds to the presence of mixed wiring with branches connected to different phases within the user's system. The actual migration volume for this user needs to be calculated separately based on the decrease in power.

[0052] Step S105: Generate an asymmetric adjustment coefficient based on the number of recurrences that still exceed the limit after the load migration indicator statistical threshold is relaxed; generate a dynamic judgment threshold based on the asymmetric adjustment coefficient and differentiated cycle parameters; and output a load adjustment command based on the load status judgment based on the dynamic judgment threshold.

[0053] In some embodiments, generating an asymmetric adjustment coefficient based on the number of recurrences exceeding the limit after the load migration identifier statistical threshold is relaxed includes: verifying historical logs for previous relaxation events to obtain relaxation event records; removing recurrence events during transient peak periods based on the relaxation event records to obtain valid recurrence identifiers; using the valid recurrence identifiers to count the number of recurrences within a preset period to determine the recurrence count value; and generating an asymmetric adjustment coefficient based on the recurrence count value according to a preset loss reduction priority grading rule.

[0054] For load migration identifiers, historical logs are checked to obtain previous relaxation event records. Records with a correlation strength ratio of 70% or higher (i.e., considered valid switching) in the load migration identifier are used to retrieve the corresponding target switch number. All event entries marked "threshold relaxation" in the historical log of that switch are then searched. A relaxation event is a record written to the historical log after the switch has not exceeded the relaxation confirmation period for a previous evaluation cycle and the relaxation decision is triggered by the load status. This means the system has determined that the switch's phase allocation has stabilized and relaxed its subsequent monitoring threshold, allowing an observation period to prevent occasional fluctuations after switching from being immediately misjudged as requiring rescheduling. The relaxation event record includes each queried relaxation event, including the time of relaxation, the temporary threshold value used after relaxation, and the corresponding observation period length. The observation period length is written at the time of relaxation and is only read in this step without recalculation. If no relaxation event has ever occurred in the historical log of this switch (such as the first completion of load migration confirmation), the relaxation event record will be empty. The asymmetric adjustment coefficient for this round will be handled according to the default case of no historical relaxation record. In the default case, the asymmetric adjustment coefficient will directly take the lowest level corresponding value. The relaxation event records are arranged from most recent to oldest according to the time of relaxation, and each record includes the start and end times of its observation period and the temporary threshold. This step will use this arranged record to unfold the observation period for each relaxation event. The arrangement order facilitates first locating the relaxation events that fall within the last twelve months, and then checking whether the load has exceeded the temporary threshold during the observation period of each event. If a new relaxation is triggered before the observation period of the previous relaxation event for the same switch has ended, the two relaxations will be recorded as independent entries. The start time of the observation period for the latter entry will be the time of the new relaxation. The temporary threshold for each record is generally one level higher than the original rated over-limit threshold. The observation period length is set according to the daily cycle length of the downstream load of the switch.

[0055] Valid recurrence identifiers are obtained by removing recurrence events during transient peak periods from the relaxed event records. For each relaxed event record, it is checked whether the switch triggers an over-limit again within its observation period. The triggering method is that the load current of the corresponding phase of the branch box exceeds the temporary threshold again at any time during the observation period. The check covers the entire evaluation cycle within the observation period, without sampling. If the over-limit trigger time falls within a preset transient peak period, such as the summer cooling load peak or the winter heating load peak, this trigger is judged as a transient recurrence and is not counted as a valid recurrence. The transient peak period is marked according to the start and end times of the seasonal peak segment in the historical load curve of the region. Such peaks are driven by temperature and are generally raised across the entire network, not caused by the failure of the phase allocation of this switch itself, so they are excluded. Valid recurrence identifiers are generated when an over-limit is triggered but the trigger time does not fall within a transient peak period. The identifier indicates the specific time of the corresponding relaxed event and the recurrence, and also includes the actual load current value at the time of triggering. For cases where the event exceeds the limit but does fall within the transient peak period, no valid recurrence marker is generated in this step. Instead, a note indicating a transient recurrence is added to the corresponding entry in the relaxed event record. This note does not affect the subsequent calculation of the asymmetric adjustment coefficient and does not trigger any on-site intervention actions. Valid recurrence markers are determined independently for each relaxed event record. If multiple triggers occur within the same relaxed event observation period, each trigger is checked separately to see if it falls within the transient peak period. Other triggers within the same observation period are not excluded simply because one has been excluded. The same applies if the first few triggers within an observation period are excluded, and only the subsequent trigger is counted as a valid recurrence; whether it is counted depends solely on whether the trigger itself falls within the peak period.

[0056] The recurrence count is determined by counting the number of recurrences within a preset period using valid recurrence identifiers. The preset period is the most recent 12 months. The number of times a valid recurrence identifier is generated within these 12 months is counted. Each valid recurrence identifier is checked to see if its corresponding recurrence time falls within this 12-month interval. Historical identifiers falling outside the interval are not included in this count. The interval boundary is calculated by extrapolating 12 calendar months backward from the current assessment time. The recurrence count is calculated using the formula C=Σδ_k, where δ_k is the indicator value for whether the k-th valid recurrence identifier falls within the preset period (1 if within the interval, 0 otherwise), and C is the recurrence count, which is a non-negative integer with no theoretical upper limit. The recurrence count is counted independently for each switch. The recurrence counts of different switches under the same branch box do not affect each other. Even if they were simultaneously switched due to the same three-phase imbalance event, each maintains its own independent recurrence history and does not borrow statistical results from the others. If no valid recurrence identifier is generated for the switch within the preset period, the recurrence count is recorded as 0, corresponding to the switch not exceeding the limit again during the observation period after the relaxation. The recurrence count is based on the specific event of "exceeding the limit after relaxation," which is distinct from the initial limit violation determination. Only the number of times a limit violation occurred after relaxation and during the observation period is included in C. For switches that have never been relaxed, C is always 0 and does not participate in the subsequent increase of the asymmetric adjustment coefficient. If multiple valid recurrence markers appear within the observation period of the same relaxation event record, the recurrence count is counted individually and accumulated into C. They are not combined into a single count because they originate from the same relaxation event, as each independent trigger reflects a genuine limit violation.

[0057] The asymmetric adjustment coefficient is generated based on the recurrence count according to a preset loss reduction priority grading rule. This rule divides the recurrence count into four levels: 0 times (Level 0), 1-2 times (Level 1), 3-4 times (Level 2), and 5 times or more (Level 3). Higher levels indicate a stronger tendency for the switch to repeatedly exceed its limits. The grading boundaries are derived from the recurrence distribution statistics accumulated over one year of operation on multiple pilot feeders using the system for this model. The asymmetric adjustment coefficient is calculated using the formula K = 1 + 0.3 × g, where g is the level value (0 to 3) corresponding to the recurrence count. The coefficient 0.3 increases K by 0.3 for each level increase, with four levels covering the adjustment range of 1.0 to 1.9. A larger K indicates a stronger tendency to tighten the subsequent judgment threshold compared to loosening it. The loss reduction priority is reflected in K's bias towards making tightening easier to trigger and loosening more difficult to trigger, thus minimizing additional losses caused by the equipment being in a state of imbalance for extended periods. The asymmetric adjustment coefficient is calculated separately for each switch. The classification and K-value calculation follow a uniform rule for all switches, without adjusting the classification boundaries based on the branch box or feeder location of the switch. When the recurrence count falls exactly at one of the two classification boundaries (e.g., exactly 2 or exactly 4 times), it is calculated according to the lower classification level; 2 times is classified as Class 1, and 4 times as Class 2. The higher classification is not applied simply because the value is larger. This rounding direction is consistent with the previous "strictly less than" approach when determining if a value exceeds the limit. When the recurrence count decreases across a classification boundary (e.g., from 3 times to 2 times, the classification level drops from 2 to 1), the asymmetric adjustment coefficient K decreases synchronously in the next statistical period without an additional delay period, consistent with the symmetrical handling principle used when extending the coefficient in subsequent periods.

[0058] In some embodiments, the step of generating a dynamic judgment threshold by converting the differentiated cycle parameter based on the asymmetric adjustment coefficient includes: converting the number of relaxation and tightening confirmation cycles based on the asymmetric adjustment coefficient to obtain the differentiated cycle parameter; analyzing the tightening process interruption records based on the differentiated cycle parameter to generate a confirmation interruption count; successively extending the tightening confirmation cycle based on the confirmation interruption count to obtain a retreat cycle parameter; and calculating the corresponding directional adjustment amplitude based on the retreat cycle parameter to generate a dynamic judgment threshold.

[0059] Differentiated cycle parameters are obtained by converting the number of confirmation cycles for tightening and loosening based on the asymmetric adjustment coefficient. This step uses the asymmetric adjustment coefficient as the sole input, scaling the original shared baseline confirmation cycle for both tightening and loosening in two opposite directions to obtain independent confirmation cycle numbers for tightening and loosening, denoted as N_tight (tightening confirmation cycle number) and N_relax (loosening confirmation cycle number), respectively, thus breaking the original symmetrical confirmation threshold between the two. The number of confirmation cycles in the tightening direction decreases proportionally with the increase of the asymmetric adjustment coefficient, allowing switches with a strong tendency to relapse to be identified as requiring rescheduling with fewer consecutive out-of-limit cycles, thus enabling faster intervention and shortening the time spent in an unbalanced state. Conversely, the number of confirmation cycles in the loosening direction increases proportionally with the increase of the asymmetric adjustment coefficient, requiring the same switch to undergo more consecutive normal cycles before being judged as stable, making it more difficult to trigger loosening. When the tightening direction shortens, a lower limit of no less than two cycles is set to avoid overly sensitive confirmation that occasional single-time exceedances might be mistaken for continuous exceedances, triggering unnecessary switching. The loosening direction only lengthens in one direction without an upper limit, reflecting a robust tendency to observe repeatedly exceeding the limit switch more closely rather than easily loosening it. Through this scaling, an asymmetric adjustment coefficient simultaneously determines the number of confirmation cycles in both directions, one short and one long, in opposite directions. The larger the coefficient, the greater the difference between the two, the more frequent the recurrence of the corresponding switch, and the more necessary it is to be constrained with a tightening-tightening-restricting approach. The confirmation cycle numbers for tightening and loosening constitute the differentiated cycle parameters, which are archived along with the corresponding switches. The tightening term is used to determine when to tighten the monitoring threshold, and the loosening term is used to determine when to loosen the monitoring threshold. The differentiated cycle parameters are obtained by converting the same asymmetric adjustment coefficient into pairs, with opposite directions and varying tightness, which is the origin of the differentiated judgment thresholds.

[0060] Based on the analysis of differentiated cycle parameters, the tightening process interruption record generates a confirmation interruption count. The current load current of the switch is checked against the assessment cycle to see if it exceeds the limit. If it does, the tightening confirmation count is incremented by 1; otherwise, the count is interrupted and reset to zero. Actual tightening is triggered when the tightening confirmation count reaches N_tight in the differentiated cycle parameter. Its cycle-by-cycle accumulation method is similar to the aforementioned accumulation of the proximity to the limit date, but the counting object is changed to the tightening confirmation process itself. If the tightening confirmation count is interrupted once or multiple times before reaching N_tight (re-counting after resetting), the assessment cycle of each interruption is recorded. The confirmation interruption count is based on the cumulative number of interruptions in that tightening confirmation process (from the first limit exceedance to the actual triggering of tightening or the end of the current observation period). The end of the observation period is determined by the absence of any new limit exceedances within 20 consecutive assessment cycles. The interruption count is independently tallied for each tightening confirmation process. Each tightening confirmation process is judged to have met the target based on the N_tight given by the differentiated cycle parameter. After the current tightening confirmation process for the same switch ends, the next new tightening confirmation process restarts the interruption count from zero, without carrying over the count from the previous process. If the tightening confirmation count successfully reaches N_tight in one go without any interruptions, the interruption count is recorded as 0. The interruption count only counts the number of times the limit is not exceeded midway between the first exceedance and the actual triggering of tightening, reflecting the degree of oscillation of the switch repeatedly entering and exiting the limit near the critical state. Once the limit exceedance stabilizes and continues, the count is 0. The interruption count is recalculated once per evaluation cycle, and the current value includes the cumulative number of interruptions in the current tightening confirmation process, the corresponding differentiated cycle parameter value, and the specific cycle number of the most recent interruption.

[0061] For example, the step of obtaining the backoff period parameter by successively extending and tightening the confirmation period based on the confirmation interruption count includes: statistically analyzing the distribution density of the confirmation interruption count within the observation period to obtain an interruption density value; determining a period extension coefficient based on the interruption density value and a density grading threshold; calculating the basic extension duration based on the period extension coefficient; and calculating the tightening confirmation period increment based on the basic extension duration to output the backoff period parameter.

[0062] The interruption density value is obtained by statistically confirming the distribution density of the interruption count within the observation period. The observation period is the most recent 30 evaluation periods. The current confirmed interruption count of the switch is read for each evaluation period, and the number of times the count increases within these 30 evaluation periods is counted as the total number of interruptions. This count covers situations spanning multiple tightening confirmation processes and is not interrupted by process boundaries. The reason for using density instead of the absolute number of interruptions is that the same number of interruptions dispersed over a long period is different from those concentrated in a short period: the former is mostly sporadic, while the latter indicates continuous oscillation of the switch near its critical state. Using the proportion of occurrences within a fixed-length observation period distinguishes between the two and makes switches with different network durations comparable. The interruption density value is calculated using the formula ρ = n_interrupt / 30, where n_interrupt is the total number of confirmed interruption count increases within the observation period, 30 is the number of evaluation periods in the observation period, and ρ is the interruption density value, ranging from 0 to 1. A larger value indicates more frequent interruptions in the tightening confirmation process per unit time. Interruption density values ​​are calculated independently for each switch, and the ρ values ​​of different switches under the same branch box do not affect each other. A ρ value close to 1 indicates that the tightening confirmation process is frequently interrupted by non-limit-crossing interruptions, and the switch load fluctuates repeatedly around the limit-crossing threshold; a ρ value close to 0 indicates that limit-crossing is clear, and tightening confirmation proceeds smoothly. If the interruption count of a switch has never increased (i.e., never been interrupted) in the last 30 evaluation periods, the interruption density value is recorded as 0, corresponding to the ideal state where the tightening confirmation process always proceeds smoothly. If the switch has undergone hardware replacement or reset during the observation period, or the interruption count has been manually reset, the decrease in value before and after the reset is not included in the increase judgment; the interruption density value only counts the number of increases accumulated after the reset.

[0063] The cycle extension coefficient is determined based on the interruption density value and the density grading threshold. The density grading threshold divides the interruption density value into three levels: below 0.2 is low density, 0.2 to 0.5 is medium density, and above 0.5 is high density. In the low density level, tightening confirmation is almost uninterrupted and does not require extending the confirmation period. In the medium and high density levels, interruptions become more frequent, and the confirmation period needs to be extended accordingly to avoid the tightening being falsely triggered during oscillations. The three level boundaries correspond to the two turning points where interruptions change from occasional to frequent. The cycle extension coefficient is set according to the level: 1.0 for low density (no extension), 1.5 for medium density, and 2.0 for high density. When the interruption density value falls exactly at the dividing point between two levels, it is classified into the lower level, consistent with the aforementioned rounding direction. The cycle extension coefficient itself does not have a fourth level. The cycle extension coefficient is determined independently for each switch. If the interruption density value of the same switch changes across the gear boundary, the cycle extension coefficient is adjusted synchronously to the value corresponding to the new gear in the next evaluation cycle. The adjustment may be an increase or a decrease, and both directions are treated equally. Different response speeds are not set because an increase is more alarming than a decrease. For switches with an interruption density value of 0 (lower limit of the low density gear), the cycle extension coefficient is 1.0, and the processing logic is consistent with the case where no interruption has occurred. It is not treated separately because the value happens to be the lower limit. The cycle extension coefficient takes three values ​​of 1.0, 1.5, and 2.0 for each gear, which are positively correlated with the interruption density value but change abruptly rather than continuously. This makes the extension of the tightening confirmation cycle step-like, avoiding continuous fine-tuning of the extension amount caused by small fluctuations in density. The step-like value also makes it easy for maintenance personnel to directly check the current oscillation level of a switch by gear.

[0064] The basic extension duration is obtained by calculating the interruption count based on the cycle extension factor. For each switch, the cycle extension factor is multiplied by the confirmation interruption count of the current tightening confirmation process. The product is the intermediate result of this step. The intermediate result itself is in units of evaluation cycles. The two values ​​before multiplication have already been determined according to their respective steps, and this step does not re-verify their source. The basic extension duration is calculated using the formula E = cycle extension factor × confirmation interruption count. E is a non-negative number (possibly containing decimals), and the unit is the evaluation cycle. When the confirmation interruption count is 0, E is 0, and no extension requirement arises. E is affected by both the gear corresponding to the cycle extension factor and the number of interruptions. The higher the number of interruptions and the higher the density gear, the larger E becomes, and the longer the tightening confirmation cycle is required to stably trigger tightening through the oscillation zone. The basic extension duration is calculated independently for each switch. Before rounding, one decimal place is retained to ensure the accuracy of the intermediate value and avoid accumulated errors caused by premature rounding. The basic extension duration itself is not directly used for the final judgment; it is only temporarily stored as a process quantity within this step. For switches with a cycle extension factor of 1.0 (low density setting) and a small number of confirmation interruptions, the E value is usually no more than a single digit. Therefore, only a slight extension of the confirmation cycle is needed to cope with the current interruption frequency. In the high density setting, coupled with the extreme case where the confirmation interruption count reaches double digits, the E value may exceed 20. In this case, the base extension duration is recorded accurately according to the formula. E is rounded to one decimal place before rounding in this step, and is only rounded up once at the final step where the incremental increase in the tightened confirmation cycle is formed. This avoids the cumulative largeness caused by successive roundings in the middle. E itself, as a process quantity, does not directly determine the judgment threshold.

[0065] The tightening confirmation cycle increment is calculated based on the base extension duration, and the backoff cycle parameter is output. First, the base extension duration is rounded up to obtain the increment of the current tightening confirmation cycle. Rounding ensures that the extension amount is not underestimated due to decimals being discarded. The backoff cycle parameter records the number of tightening confirmation cycles after the switch adjustment, and its value is the sum of the original tightening confirmation cycle number and the current increment. This value is directly applied as the actual number of cycles used for tightening confirmation in this round of switch tightening confirmation. This increasingly longer and more oscillating processing borrows from the backoff approach in communication transmission, where repeated conflicts lead to progressively longer retry intervals: when the load of a switch fluctuates around the over-limit threshold and tightening confirmation is repeatedly interrupted, using a fixed confirmation cycle would cause the judgment to repeatedly jump between tightening and resuming, resulting in unnecessary repeated phase switching. By progressively increasing the confirmation cycle with the frequency of interruptions, it is equivalent to requiring the over-limit to last longer and pass through the entire oscillation zone before being accepted, thereby stabilizing the switch in a critical state and preventing it from being repeatedly scheduled. If the tightening confirmation process proceeds smoothly and without interruption (basic extension time is 0), the increment is 0, the backoff period parameter is equal to the original tightening confirmation period number, and no additional extension is made. The corresponding switch does not require backoff for the time being. The backoff period parameter is maintained independently for each switch, with no upper limit on its value. If a switch experiences frequent interruptions or a persistently high basic extension time, the parameter will accumulate and increase sequentially. Backoff only applies to the tightening direction. The longer the tightening confirmation period, the longer the over-limit condition needs to be met before it is confirmed as requiring rescheduling, thus suppressing ineffective repeated commutation on oscillating switches.

[0066] The corresponding directional adjustment range is calculated based on the backoff cycle parameter to generate a dynamic judgment threshold. The tightening directional adjustment range corresponding to the backoff cycle parameter is taken as the difference between this parameter and the original N_tight in the differentiated cycle parameter. The larger the difference, the longer the confirmation period is required for this tightening judgment relative to the benchmark, meaning the judgment standard for the tightening direction is relatively relaxed. This difference itself is also a direct reference for maintenance personnel to judge whether a switch is in a state of repeated oscillation. Switches with a persistently large difference often require maintenance personnel to check the wiring condition of the line itself on-site. The relaxed directional adjustment range is directly taken as N_relax in the differentiated cycle parameter and is not affected by the current backoff cycle parameter, because the caution of the relaxed direction has already been reflected once by the asymmetric adjustment coefficient K when converting the differentiated cycle parameter, and does not need to be superimposed again in this step. The dynamic judgment threshold includes two values: the number of currently effective tightening confirmation cycles (i.e., the backoff cycle parameter) and the number of relaxation confirmation cycles (i.e., N_relax). These two values ​​serve as thresholds for determining when to tighten and when to relax the monitoring limits, respectively. The confirmation cycle for the tightening direction is dynamically extended by the backoff mechanism based on the interruption situation, while the confirmation cycle for the relaxation direction is fixed and converted from the asymmetric adjustment coefficient. The two directions are separate and independent of each other. For this switch being included in the evaluation of this method for the first time, and for which the differentiated cycle parameter and backoff cycle parameter have never been calculated before, the dynamic judgment threshold directly uses the benchmark confirmation cycle of 6 as the initial value for both the tightening and relaxation directions. After at least one complete tightening or relaxation judgment is completed, it will switch to the differentiated calculation result of this step. After the switch is switched, it will be calculated independently according to its own recurrence history and will no longer share the benchmark value.

[0067] Load status is determined based on dynamic judgment thresholds, and load adjustment commands are output. For each evaluation cycle, the current load current of the switch is compared with the tightening direction judgment criteria in the dynamic judgment thresholds. If the current exceeds the limit continuously for a certain number of cycles, a tightening judgment is triggered. The tightening judgment result indicates that the three-phase imbalance of the switch is persistent and requires re-entry into the heavy load phase analysis stage. Upon re-entry, the timing information of this judgment is carried for reference by that stage. If the same switch does not exceed the limit continuously for a certain number of cycles corresponding to the easing direction judgment criteria in the dynamic judgment thresholds, a easing judgment is triggered. The judgment result indicates that the current phase distribution of the switch has stabilized and no new scheduling action is needed in this round. Simultaneously, this easing event is written to the historical log for review and recurrence in subsequent evaluation cycles. A load adjustment command is generated when a tightening decision is triggered. It contains a record with the switch number, trigger time, and a rescheduling command word. The command word is set to "rescheduling," thus initiating a new round of the entire process starting from the identification of structural imbalance sections. Similarly, a load adjustment command is generated when a relaxation decision is triggered. This record contains the switch number, trigger time, and a "maintain status" command word. The command word is set to "maintain status," thus preserving the current phase allocation of the switch and not initiating a new round of scheduling. In other words, the load adjustment command is ultimately a decision record with the switch number as the target, "rescheduling" or "maintain status" as the command word, and a trigger time. The former drives the system to re-execute the entire commutation process, while the latter prevents the system from switching the switch in this round. When the command word is "rescheduling," the command locates the corresponding double-throw commutator switch according to the switch number and sends it to the heavy-load phase analysis to await commutation. When the command word is "maintain status," it only keeps the double-throw switch contacts in their current closed phase without operation.

[0068] To implement the three-phase imbalance adjustment method for a low-voltage cable branch box corresponding to the above method embodiment, in order to achieve the corresponding functions and technical effects. See also Figure 4 , Figure 4 This paper shows a structural block diagram of a three-phase imbalance adjustment system 400 for a low-voltage cable branch box according to an embodiment of this application, including: The status identification unit 401 is used to acquire three-phase current and phase-to-phase voltage through the in-box current transformer to obtain operating parameters, and to determine the structural imbalance section based on the operating parameters by performing no-load segment deviation analysis and zero-load baseline drift analysis. The identifier generation unit 402 is used to analyze the characteristics of the heavy-load branch in the structural imbalance section to generate a heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes of the same feeder based on the heavy-load phase identifier to obtain the target phase conflict identifier, and select non-conflicting light-load phases to form a candidate switch identifier group according to the target phase conflict identifier. The target selection unit 403 is used to calculate the over-limit duration ratio through the candidate switch identification group to obtain the over-limit day proximity, and to calculate the screening tolerance threshold according to the urgency level based on the over-limit day proximity, and to evaluate the operation margin of the upper-level protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group to obtain the target switch identification. The switching execution unit 404 is used to determine the action parameters based on the target switch identifier to form switching control parameters, transmit the switching control parameters to the on / off state of the switch output contact, and analyze the correlation of the user-side electricity meter's electricity consumption characteristics based on the on / off state of the contact to obtain a load migration identifier. The instruction output unit 405 is used to generate an asymmetric adjustment coefficient based on the number of recurrences that still exceed the limit after the load migration identifier statistical threshold is relaxed, to generate a dynamic judgment threshold based on the asymmetric adjustment coefficient to convert the differentiated cycle parameter, and to output a load adjustment instruction based on the dynamic judgment threshold to determine the load status.

[0069] The aforementioned three-phase imbalance adjustment system 400 for a low-voltage cable branch box can implement a three-phase imbalance adjustment method for a low-voltage cable branch box as described in the above-described method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining contents of this application embodiment can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0070] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0071] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for adjusting three-phase imbalance in a low-voltage cable branch box, characterized in that, include: Operating parameters are obtained by collecting three-phase current and phase-to-phase voltage through the in-box current transformer. Based on the operating parameters, the structural imbalance section is determined by analyzing the no-load segment deviation and zero-load baseline drift. For the structural imbalance section, analyze the characteristics of the heavy-load branch to generate a heavy-load phase identifier. Based on the heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes on the same feeder to obtain the target phase conflict identifier. According to the target phase conflict identifier, select non-conflicting light-load phases to form a candidate switch identifier group. The over-limit duration ratio is calculated by the candidate switch identification group to obtain the over-limit day proximity. The over-limit day proximity is graded according to the urgency to obtain the screening tolerance threshold. The target switch identification is obtained by evaluating the operation margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group. Based on the target switch identifier, the action parameters are determined to form switching control parameters. The switching control parameters are transmitted to the on / off state of the switch output contacts. Based on the on / off state of the contacts, the correlation of the electricity consumption characteristics of the user-side electricity meter is analyzed to obtain the load migration identifier. Based on the number of recurrences that still exceed the limit after the load migration identifier statistical threshold is relaxed, an asymmetric adjustment coefficient is generated. Based on the asymmetric adjustment coefficient, a differentiated periodic parameter is converted to generate a dynamic judgment threshold. Based on the dynamic judgment threshold, the load status is judged and a load adjustment command is output.

2. The method according to claim 1, characterized in that, The determination of structural imbalance sections based on the above operating parameters through unloaded segment deviation analysis and zero-load baseline drift analysis includes: The no-load segment data is obtained by screening periods when the load current is lower than the no-load threshold from the operating parameters. The no-load deviation fluctuation value is obtained by statistically analyzing the three-phase current deviation fluctuation based on the no-load segment data. Based on the comparison of the no-load deviation fluctuation value with the preset zero-load baseline interval, a baseline drift is generated; Based on the baseline drift amount, if it exceeds the allowable range of sensing error, a structural imbalance section is determined.

3. The method according to claim 1, characterized in that, The step of identifying the target phase conflict identifier based on the heavy-load phase identifier to determine the scheduling conflict status of adjacent branch boxes on the same feeder includes: The communication link between adjacent boxes is obtained by querying the communication address of the adjacent branch box on the same feeder for the heavy-load phase identifier; Based on the neighbor box communication link, the pending phase switching instruction is read to generate the neighbor box suspension intention; Based on the neighbor box suspension intention, candidate switching targets are identified and concurrent conflict identifiers are generated concurrently; The target conflict identifier is obtained by checking the preset conflict conditions based on the concurrent conflict identifier.

4. The method according to claim 1, characterized in that, The process of obtaining the target switch identifier by evaluating the operating margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identifier group includes: The protection setting value is obtained by verifying the protection setting position of each candidate switch according to the candidate switch identification group; Based on the protection setting value, the fault identification margin is determined by calculating the fault current margin after transfer. Based on the fault identification margin ratio, a protection risk identifier is generated when the screening tolerance threshold is exceeded and a reliable action limit is detected. Based on the protection risk identifier, faults are eliminated, abnormal candidate switches are identified, and the target switch identifier is obtained.

5. The method according to claim 1, characterized in that, The process of analyzing the correlation between the user-side electricity meter's electricity consumption characteristics and the on / off state of the contact points to obtain load migration identifiers includes: A target user list is established by parsing the corresponding downstream branch user account number list from the on / off state of the contact points. Based on the target user list, retrieve the power curves of the electricity meters before and after the switchover to form the user power curves; Based on the user power curve, the power mutation identifier is obtained by verifying that the sudden change at the switching time is in line with expectations. The load migration identifier is obtained by evaluating the branch load association strength based on the power mutation identifier.

6. The method according to claim 1, characterized in that, The step of generating an asymmetric adjustment coefficient based on the number of relapses that still exceed the limit after relaxing the statistical threshold for load migration indicators includes: For the load migration identifier, historical logs were checked to obtain previous relaxation events and relaxation event records; Based on the relaxed event record, recurrence events during transient peak periods are removed to obtain effective recurrence markers; The number of recurrences within a preset period is determined by using the effective recurrence markers. Based on the recurrence count, an asymmetric adjustment coefficient is generated according to a pre-set priority classification rule for loss reduction.

7. The method according to claim 1, characterized in that, The step of generating a dynamic judgment threshold based on the differential periodic parameters converted from the asymmetric adjustment coefficient includes: Based on the aforementioned asymmetric adjustment coefficient, the number of relaxation and tightening confirmation cycles is converted to obtain differentiated cycle parameters; Based on the analysis of the differentiated periodic parameters, the tightening process interruption record is used to generate a confirmed interruption count; The backoff period parameter is obtained by successively extending and tightening the confirmation period based on the confirmed interruption count; The dynamic judgment threshold is generated by calculating the corresponding directional adjustment range based on the retreat cycle parameters.

8. The method according to claim 2, characterized in that, The determination of structural imbalance segments based on the baseline drift exceeding the permissible range of sensing error includes: A preliminary threshold indication is generated for the baseline drift amount, based on a preset error tolerance range. Based on the initial limit violation indicator, the number of repeated limit violations in consecutive periods is counted to obtain a continuous limit violation count; Based on the continuous over-limit counting confirmation that a preset continuous threshold has been reached, a drift confirmation flag is generated. The drift confirmation flag is used to exclude occasional false alarms and identify structural imbalance sections.

9. The method according to claim 7, characterized in that, The backoff period parameter is obtained by successively extending and tightening the confirmation period based on the confirmation interruption count, including: The interruption density value is obtained by statistically analyzing the distribution density of the confirmed interruption counts within the observation period; The period extension coefficient is determined based on the interruption density value and the density grading threshold. The basic extended duration is obtained by calculating the confirmed interruption count based on the cycle extension coefficient. Based on the aforementioned basic extended duration, the tightening confirmation cycle is calculated and the retreat cycle parameter is output with an incremental increase.

10. A three-phase imbalance adjustment system for a low-voltage cable branch box, characterized in that, include: The status identification unit is used to acquire three-phase current and phase-to-phase voltage through the in-box current transformer to obtain operating parameters, and to determine the structural imbalance section based on the operating parameters by performing no-load segment deviation analysis and zero-load baseline drift analysis. The identifier generation unit is used to analyze the characteristics of the heavy-load branch in the structural imbalance section to generate a heavy-load phase identifier, identify the scheduling conflict status of adjacent branch boxes of the same feeder based on the heavy-load phase identifier to obtain the target phase conflict identifier, and select non-conflicting light-load phases to form a candidate switch identifier group according to the target phase conflict identifier. The target selection unit is used to calculate the over-limit duration ratio through the candidate switch identification group to obtain the over-limit day proximity, and to calculate the screening tolerance threshold according to the urgency level based on the over-limit day proximity, and to evaluate the operation margin of the upstream protection device of each candidate switch based on the screening tolerance threshold and the candidate switch identification group to obtain the target switch identification. The switching execution unit is used to determine the action parameters based on the target switch identifier to form switching control parameters, transmit the switching control parameters to the on / off state of the switch output contact, and analyze the correlation of the user-side electricity meter's electricity consumption characteristics based on the on / off state of the contact to obtain a load migration identifier. The instruction output unit is used to generate an asymmetric adjustment coefficient based on the number of recurrences that still exceed the limit after the load migration identifier statistical threshold is relaxed, to generate a dynamic judgment threshold by converting the differentiated cycle parameter based on the asymmetric adjustment coefficient, and to output a load adjustment instruction based on the load status judgment based on the dynamic judgment threshold.

Images