Power communication multi-band signal intelligent switching transmission system

Abstract

The application discloses a power communication multi-frequency band signal intelligent switching transmission system and particularly relates to the field of power line communication transmission, and is used for solving the problem of slow response of a short-time link fluctuation to multi-frequency band switching in a power distribution line medium. The power communication multi-frequency band signal intelligent switching transmission system generates a state sequence by sending a detection frame according to a unified time scale through a short window weaving sequence module, scans the sequence to identify fluctuation boundaries, outputs a candidate frequency band set and a residence marker through a stable scale plotting module, generates a switching holding coefficient to execute frequency band switching through a sequence position joint judgment module, checks a table relationship in a switching window and corrects and writes back the sequence through a table writing back module, and iteratively updates a rule and fine-tunes a detection solidified strategy in an idle period through a strategy convergence module, so that the frequency band selection is timely and stable.

Description

Technical Field

[0001] This invention relates to the field of power line communication transmission, and more specifically, to a power communication multi-band signal intelligent switching transmission system. Background Technology

[0002] In power communication transmission systems, power distribution lines are typically used as the communication medium. Multiple available frequency bands are deployed within the same medium, carrying control messages and service data through continuous detection and optimal switching. The field environment exhibits diverse interference patterns, and link quality fluctuates significantly within short periods. Multi-band switching needs to be judged and executed within short pauses, avoiding prolonged stays on unfavorable frequency bands while preventing transmission jitter caused by frequent switching. To achieve stable transmission, frequency band assessment often relies on windowed observation and dwell strategies. When multiple source states alternate within a short period, the disconnect between assessment and decision-making is amplified, making the timing and duration of frequency band switching highly sensitive. When services primarily consist of short messages, any delay or misjudgment will accumulate as retransmissions and waiting at the upper layers, ultimately resulting in peak end-to-end latency and decreased link availability, thus compressing the value of intelligent multi-band switching. Within the same power line medium, enabling frequency band selection to respond more directly and promptly to short-term changes becomes a prerequisite for stable multi-band transmission.

[0003] However, the frequency band assessment and retention strategies are insufficient in identifying short-term abrupt changes. Multi-band handover within the same medium is insensitive to short-window-level link fluctuations, resulting in excessively long retention periods and handover jitter. This manifests as the actual link entering an unfavorable segment while the assessment window still indicates availability, leading to a delayed handover action. The handover is triggered only after the assessment result turns unfavorable, followed by a short period of improvement, triggering a handover back. This is accompanied by dense retransmission of short packets and intermittent spikes in end-to-end latency, disrupting service continuity. Maintenance-side positioning granularity is insufficient, making scenario reproduction difficult. The core technical problem lies in the lack of fine-grained discrimination and stable retention strategies for multi-band intelligent handover within the same power line medium in response to short-window-level link fluctuations. This results in asynchronous assessment results and execution actions, leading to untimely and unreliable frequency band selection.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent switching transmission system for multi-band power communication signals. This system generates a frequency band short-window state sequence by sending short probe frames under a unified time scale using a short-window sequencing module. A stabilization plotting module scans the sequence to identify fluctuation boundaries and outputs a candidate frequency band set and a retention flag. A sequence position judgment module generates switching holding coefficients to execute frequency band switching. A table write-back module verifies the table relationship within the switching window and writes back the correction sequence. A strategy convergence module iteratively updates the rules and fine-tunes the probe-fixed strategy during idle periods, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Short window sequencing module: Under a unified time scale, short probe frames are sent sequentially to each frequency band of the same power line medium and the receipt time and error frame prompts are read. The frequency band short window status sequence is generated in chronological order.

[0008] The continuity and stability plotting module identifies fluctuation boundaries and forms continuity indicators on the short-window state sequence of frequency bands, and outputs the candidate frequency band set and the retention mark of the current frequency band, which together serve as the input for the merging decision;

[0009] Sequence merging module: At the merge decision point, a handover holding coefficient is generated based on the short window discrimination result. When the coefficient points to a handover tendency, the target frequency band is selected from the candidate frequency bands and the handover is performed. When the coefficient points to a retention tendency, the retention is maintained.

[0010] Table writing-back module: Collects the table relationship and short window trajectory of the frequency band before and after the handover in the handover window, completes the handover result verification based on the table relationship, performs review and correction when the verification display falls back, and writes the correction result back to the frequency band short window status sequence;

[0011] Strategy convergence module: Based on the short-window state sequence of the frequency band after write-back, it iteratively updates the candidate patterns and retention markers, initiates fine-tuning probes during idle periods to correct long-term mismatches, and generates stable handover strategies and execution records.

[0012] Furthermore, the short window sequencing module performs time anchor alignment on the acknowledgment landing points of different frequency bands. When the time interval of adjacent probes deviates from the beat tolerance band, a beat offset mark is added. The rising edge of the frame error prompt and the spike of the acknowledgment time are registered as event scales. The order and interval of occurrence are recorded with the original time stamp, forming a structured record of short window-slot-event.

[0013] Furthermore, the stabilization plotting module scans the error frame prompt bit and the acknowledgment time trajectory along the time axis on the short window state sequence of the frequency band. When the error frame prompt changes from closed to active or the acknowledgment time exceeds the convergence band, the fluctuation boundary is calibrated, and the static stable segment and the disturbance segment are segmented according to the boundary. The continuity indication is calculated based on the segment splicing relationship.

[0014] Furthermore, the continuity mapping module selects frequency bands that meet the static stability requirements according to the continuity indication to form a candidate frequency band set, and writes the static stability connectivity of the current carrying frequency band into the retention flag, and uses the candidate frequency band set and the retention flag as input for subsequent merging and discrimination.

[0015] Furthermore, at the merging and discrimination stage, the sequence position merging module generates a rhythmic sync trajectory index and a static stable duration for each candidate frequency band. The rhythmic sync trajectory index first reads the event scale and projects it onto the detection beat slot to form an occupancy map. Then, it completes the rule mapping by the occupancy continuity of adjacent slots and the overlap of occupancy with the beat. The static stable duration first locates the static stable connected interval that can cross the switching window based on the distribution of static stable segments, and then takes the shortest acceptance time of the corresponding connected interval as the duration.

[0016] Furthermore, the sequence judgment module inputs the rhythm coincidence trajectory index and the static steady duration into the sequence segmentation tree learner according to the order of judgment timing and subsequent acceptance. The splitting criteria are the consistency of rhythm coincidence area and free area, sufficient acceptance area and insufficient acceptance area and their change direction. The leaf node outputs the confidence prototype of switching tendency or retention tendency, which is calibrated and mapped to the switching holding coefficient. When it points to the switching tendency, the target frequency band is selected from the candidate frequency band set and the switching is performed. When it points to the retention tendency, the state pointed to by the retention mark is maintained.

[0017] Furthermore, the table write-back module opens the switching window when the switching action is triggered, and synchronously collects the acknowledgment sequence and error frame prompt sequence of the frequency band before switching and the target frequency band. The table relationship is established with a unified time anchor. The table relationship consists of message arrival order alignment, acknowledgment time landing point alignment, and error frame prompt edge alignment. The consistency of the three is used as the basis for verifying the switching result.

[0018] Furthermore, when the table write-back module shows a decline in performance, it initiates a review and correction process. It checks the detection beat slot and event scale in reverse order along the switching window. When a phase misalignment between the beat and the event is detected, a beat correction rule is registered. When a table landing point offset is detected, a landing point correction rule is registered. The beat correction rule, the landing point correction rule, and the short window trajectory observed in the switching window are all written back to the frequency band short window state sequence to unify the data caliber.

[0019] Furthermore, during idle periods, the strategy convergence module reads the frequency band short-window state sequence that has been corrected by write-back, and iteratively updates the candidate patterns and retention markers. The candidate patterns are revised and incorporated into the boundary according to the static stable connectivity pattern and the disturbance rhythm. The retention markers are revised according to the switching window verification statistics and the attention and retention time limit of the boundary. The newly revised beat correction rules and landing point correction rules are verified by triggering micro-debugging with small flow. When the feedback is consistent, it is solidified into the strategy record. When it is inconsistent, it is backfilled into the rule revision pool for merging. Finally, a stable switching strategy and execution record are formed, and the switching time, target frequency band, verification result and correction rule number are saved in chronological order.

[0020] The technical effects and advantages of the intelligent switching transmission system for multi-band signals in power communication according to the present invention are as follows:

[0021] This invention uses power lines as the sole medium. It first uses short probes to establish a short window state, then compresses two key information categories—rhythm and acceptance—into a single criterion. Switching actions are checked against a table within the window and written back immediately. During idle periods, small-volume trials are conducted to ensure the strategy converges according to field behavior. As a result, frequency band selection responds faster and more stably to short-term fluctuations, significantly reduces excessive dwell time and back-and-forth switching, synchronously converges delay spikes and retransmission accumulation, maintains continuous control messages and service transmission, aligns execution with field realities, makes recording and correction rules traceable, and provides intuitive maintenance and positioning. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the structure of a multi-band intelligent switching transmission system for power communication according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1: Figure 1 The present invention discloses an intelligent switching transmission system for multi-band signals in power communication, comprising:

[0025] Short window sequencing module: Under a unified time scale, short probe frames are sent sequentially to each frequency band of the same power line medium, and the receipt time and error frame prompts are read. The frequency band short window status sequence is generated in chronological order.

[0026] The continuity and stability plotting module identifies fluctuation boundaries and forms continuity indicators on the short-window state sequence of frequency bands, and outputs the candidate frequency band set and the retention mark of the current frequency band. The two together serve as inputs for the merging decision.

[0027] Sequence merging module: At the merge decision point, a handover holding coefficient is generated based on the short window discrimination result. When the coefficient points to a handover tendency, the target frequency band is selected from the candidate frequency bands and the handover is performed. When it points to a retention tendency, the retention is maintained.

[0028] Table writing-back module: Collects the table relationship and short window trajectory of the frequency band before and after the handover in the handover window, completes the handover result verification based on the table relationship, performs review and correction when the verification display falls back, and writes the correction result back to the frequency band short window status sequence.

[0029] Strategy convergence module: Based on the short-window state sequence of the frequency band after write-back, it iteratively updates the candidate patterns and retention markers, initiates fine-tuning probes during idle periods to correct long-term mismatches, and generates stable handover strategies and execution records.

[0030] In the field of power communication, when power distribution lines serve as the communication medium for multi-band signal transmission, they face variable interference from the field environment. This interference causes drastic fluctuations in link quality within a short period, thus affecting the stable transmission of control messages and service data. Traditional multi-band switching mechanisms rely on windowed observation, but they are slow to respond to short-term fluctuations, easily leading to excessively long dwell times or frequent backswitches, which in turn cause end-to-end delay spikes and a decrease in link availability. This invention introduces a fine-grained detection and serialized recording mechanism through a short-window sequencing module to ensure timely capture of short-window-level changes in frequency band status.

[0031] However, existing frequency band assessment strategies are difficult to achieve synchronous discrimination and stable holding when faced with short-term interference within the same power line medium, resulting in delayed or jittery handover actions. This module accurately captures the dynamic changes in acknowledgment time and frame error prompts through ordered detection and structured sequence generation under a unified time scale, enabling direct response to link fluctuations.

[0032] The specific processing logic of the short window weaving module:

[0033] 1-1) Send short probe frames and read data.

[0034] Under a unified timescale, short probe frames are sequentially transmitted to various frequency bands of the same power line medium according to a preset frequency band order. These short probe frames are designed as small, fixed-length data packets used to test link quality. Immediately after transmission, the acknowledgment time (i.e., the round-trip time from transmission to receipt) and error frame indication (a binary flag indicating whether a frame error or loss occurred) for each short probe frame are read. This read data is written to a record at a unified timescale, ensuring that data from all frequency bands are aligned on the timeline. This ordered transmission and reading process ensures data timing consistency, facilitating subsequent analysis of link fluctuations and avoiding time deviations between multiple frequency bands.

[0035] Under a unified time scale, short probe frames are sent one by one from the first to the last in the frequency band sequence (where the frequency band index represents the current frequency band), the acknowledgment time and error frame prompt (zero for no error, one for error) are recorded, and written to the corresponding time scale point, where the probe sequence number indicates the current probe sequence position.

[0036] 1-2) Divide the short window and create fields.

[0037] Based on the acknowledgment time and error frame indication, short windows are segmented according to the detection cycle (defined as a fixed-interval detection period, e.g., once every 50 milliseconds), with each short window corresponding to a continuous detection cycle segment. Three types of fields are established for each short window: cycle slot (representing the position sequence of the detection cycle within the short window), acknowledgment time trajectory (a continuous sequence of acknowledgment time values), and error frame indication bit (a bit sequence indicating error frames). These fields together form the frequency band short window state sequence, where the short window index indicates the current short window order. This field-based approach organizes the raw data into a structured form, facilitating the tracking of short-term changes and improving the analyzability of the sequence.

[0038] The detection clock is determined based on the link fluctuation period prediction. For each short window (the short window number indicates the current short window), the clock slots are arrayed (from the first slot to the last slot, where the number of slots indicates the total number of slots), the acknowledgment time trajectory is arrayed from the first acknowledgment time to the last acknowledgment time, and the error frame indication bits are arrayed from the first error frame indication to the last error frame indication. These are integrated into a frequency band short window state sequence, including the clock slots, acknowledgment time trajectory, and error frame indication bits.

[0039] 1-3) Execute time anchor alignment.

[0040] For the receipt points of different frequency bands in the formed frequency band short window state sequence, a time anchor alignment operation is applied to unify the receipt time of all frequency bands to the same reference time scale. The alignment mechanism corrects the time distance deviation between detections, ensures that multi-band data are compared at a uniform scale, and avoids misjudgments caused by offset.

[0041] The time anchor is defined as a reference point for a unified time scale. For a specific frequency band, the deviation between the retrieval point and the time anchor is calculated. The value is the absolute value of the retrieval point minus the time anchor. If the deviation exceeds the clock tolerance band (the tolerance band is set to negative - ∈, ∈ through simulated interference environment testing, where ∈ = 10ms), a clock offset mark is made in the corresponding slot (the value is the direction and magnitude of the deviation, such as positive deviation + δ).

[0042] 1-4) Register event scale.

[0043] Based on the above, the rising edge of the error frame indication bit (the transition point from zero to one) and the spike in the acknowledgment time (the point where the acknowledgment time increases sharply) are scanned, and these points are registered as event scales. The occurrence order (sequential position) and interval (time difference between adjacent events) of the event scales are recorded using the original time scale, thus forming a structured record of short window-slot-event.

[0044] The rising edge detection of error frame prompts registers the first type of event scale when the current error frame prompt bit is 1 and the previous error frame prompt bit is 0. The acknowledgment time abruptness judgment uses abrupt change detection, the value of which is the current acknowledgment time minus the previous acknowledgment time. If the absolute value of this value is greater than a threshold (the threshold is set as a multiple of the average acknowledgment time through link noise distribution analysis), the second type of event scale is registered. The event scale includes the first type and the second type, the order is the index of the event in the sequence, the interval is the difference between the next time scale and the previous time scale, and it is integrated into the structured record.

[0045] Short probe frames are sent sequentially by frequency band under a unified time scale. The acknowledgment time and error frame prompt are read and written to the same time scale landing point. Short windows are divided according to the probe beat. Three types of fields are established for each short window: beat slot, acknowledgment time trajectory, and error frame prompt bit, forming a frequency band short window state sequence. Time anchor alignment is performed on the acknowledgment landing points of different frequency bands. When the time interval of adjacent probes deviates from the beat tolerance band, a beat offset mark is marked. The rising edge of the error frame prompt and the protrusion of the acknowledgment time are registered as event scales. The order and interval of occurrence are recorded with the original time scale, forming a structured record of short window-slot-event.

[0046] Through the above processing, the short-window sequencing module achieves accurate serialization and capture of short-term link fluctuations in multiple frequency bands of the power line medium, ensuring that the frequency band short-window state sequence serves as a reliable input for subsequent modules.

[0047] The robust plotting module takes the frequency band short window state sequence as input, scans the error frame prompt bit and the receipt time trajectory along the time axis to identify fluctuation boundaries and quantify continuity indication, ensuring robust plotting of frequency band states in a power distribution environment with variable interference, and outputs a candidate frequency band set and retention mark as the core basis for merging decision.

[0048] The specific processing logic of the continuous stabilization plotting module:

[0049] 2-1) Scan and calibrate the undulation boundaries.

[0050] The input is a short-window state sequence within the frequency band. The system scans the error frame alert bits and retrieval time trajectories point-by-point along the time axis, detecting transition points where error frame alerts change from closed to active, or where retrieval time values ​​exceed the upper limit of a preset convergence band. These points are designated as fluctuation boundaries. The convergence band is defined as the range of retrieval time fluctuations in a stable state; for example, the lower and upper limits of retrieval times within past observation periods are used as boundary values. This meticulous scanning process captures minute deviations in link quality, ensuring that boundary calibration reflects the actual onset of interference and avoiding omissions from coarse-grained observations.

[0051] For each short window, the position where the current bit value is active and the previous bit value is closed is identified as a boundary. This detection highlights the suddenness of frame error events. For the receipt time trajectory sequence, if the current receipt time value is greater than the upper limit of the convergence band (the upper limit is set as a multiple of the average receipt time by simulating the receipt distribution under a stable link environment), it is marked as another boundary, emphasizing the anomaly of time extension. All fluctuation boundaries are recorded in time-stamped order to facilitate the continuous processing of subsequent segment division.

[0052] 2-2) Segmentation.

[0053] Based on the calibrated fluctuation boundaries, the frequency band short-window state sequence is divided along the time axis into stable segments (continuous segments without fluctuation boundaries) and disturbed segments (continuous segments containing fluctuation boundaries). Each segment records its start time stamp, end time stamp, and length. This segmentation method decomposes the sequence into independent units, facilitating the quantification of the distribution characteristics of different states and improving the accuracy of interference mode identification.

[0054] Starting from the beginning of the sequence, the current segment ends and a new segment begins when a fluctuation boundary is encountered; if there is no boundary, it is a static stable segment, and its length is equal to the end time index minus the start time index. This calculation highlights the persistence of the stable period; if there is a boundary, it is a perturbation segment, and its length is equal to the end time index minus the start time index. This emphasizes the temporal span of the disturbance; all segments are stored sequentially as a segment set, providing a structured basis for continuous computation.

[0055] 2-3) Calculate the continuity indicator.

[0056] By leveraging the splicing relationships within a fragment set, a continuity indicator is calculated. This indicator quantifies overall stability through the connectivity of statically stable fragments and the scrambling degree of perturbed fragments. The calculation integrates the duration and quantity ratio of fragments to ensure that the indicator value comprehensively reflects the link's quality profile and avoids biases from single-dimensional assessments.

[0057] First, the static stability connectivity is calculated, which is the sum of the lengths of all static stability segments divided by the total duration of the sequence. This ratio highlights the proportion of stable periods. Second, the perturbation scrambling degree is calculated, which is the number of perturbation segments divided by the sum of the number of static stability segments and the number of perturbation segments. This calculation quantifies the density of interference. The continuity indicator is equal to the static stability connectivity multiplied by one minus the perturbation scrambling degree, ranging from zero to one. A higher value indicates more continuous static stability and less perturbation, thus providing a reliable numerical basis for frequency band selection.

[0058] 2-4) Output candidate frequency band set and retention flag.

[0059] Based on the continuity indication, frequency bands whose continuity indication exceeds the static stability threshold are selected to form a candidate frequency band set. Simultaneously, a retention flag is written to the continuity indication of the currently carried frequency band (the value is the continuity indication; if it is higher than the static stability threshold, it is marked as retention priority). The static stability threshold is determined, for example, through field interference testing, and is the lowest continuity indication value that can maintain service continuity. This output mechanism prioritizes high-stability frequency bands, ensuring optimized decision input and reducing the risk of invalid handovers.

[0060] The continuity indicator of all frequency bands is traversed. If the continuity indicator is greater than the static stability threshold, it is added to the candidate frequency band set. This screening highlights the priority of high-quality frequency bands. For the current frequency band, the retention flag is equal to the continuity indicator and is accompanied by a priority label, which facilitates the subsequent judgment of the necessity of retention. The candidate frequency band set and retention flag are output for direct use in the merging decision.

[0061] Scan the error frame prompt bit and the acknowledgment time trajectory along the time axis on the short window state sequence of the frequency band. When the error frame prompt changes from closed to active or the acknowledgment time exceeds the convergence band, mark the fluctuation boundary. Divide the statically stable segment and the disturbed segment according to the boundary, and calculate the continuity indicator based on the segment splicing relationship. This indicator reflects both the statically stable connectivity and the disturbance mixing degree. Based on this, select the frequency bands that meet the static stability requirements to form a candidate frequency band set, and write the retention mark according to the statically stable connectivity of the current carrying frequency band. Use the two results as input for subsequent merging and discrimination.

[0062] Through the above processing, the continuity mapping module completes the boundary mapping and continuity quantization of the short window state sequence of the frequency band, identifies short-term fluctuation patterns, and ensures that the output of the candidate frequency band set and the retention mark improves the robustness of the handover decision and reduces transmission jitter and latency spikes.

[0063] Based on the candidate frequency band set and retention markers output by the continuous stability plotting module, the frequency band options that meet the static stability requirements are accurately located through fluctuation boundary identification and continuity indication quantification, and the retention priority of the currently carried frequency band is marked, thus providing a robust input basis for merging decisions. The sequence merging module takes this as the starting point, generates two key quantities at the merging discrimination point and inputs them into the sequence segment tree learner to form the switching holding coefficient. The sequence logic is used to fuse the interference rhythm and static stability, ensuring the mapping of decision tendency under short-term fluctuations of the power line medium, and laying the foundation for an immediate response mechanism for the switching execution and verification of the subsequent table write-back module.

[0064] The specific processing logic of the sequence number comparison module:

[0065] 3-1) Generate rhythmic trajectory index.

[0066] The input consists of each frequency band in the candidate frequency band set. Event scales are read and projected onto the detector clock slots to form an occupancy map, which records the distribution of events within the slots. Next, the occupancy continuity (the degree of continuity of event occupancy between adjacent slots) and the overlap tightness of occupancy with the clock cycle are calculated. A rhythmic consistency trajectory index is generated through rule mapping. The closer the index value is to one end, the more stable the overlap between the interference mode and the detector clock cycle, facilitating the prediction of switching timing. This projection and mapping process dynamically transforms events into quantitative indicators, ensuring fine-grained assessment of interference predictability and avoiding blind decision-making.

[0067] For each frequency band in the candidate frequency band set, first extract the event scale (including sequence and interval), and project each event scale onto the specific position of the corresponding detection beat slot. For example, if the event sequence falls into a certain slot range, mark the slot as occupied to form an occupancy map, including the occupancy status from the first slot to the last slot, where the total number of slots represents the total number of slots.

[0068] Then, the occupancy continuity of adjacent slots is calculated, which is the number of adjacent occupied slot pairs divided by the total number of adjacent slot pairs. This ratio reflects the density of continuous occupancy. At the same time, the overlap tightness of occupancy with the beat is calculated, which is the number of occupied slots multiplied by the average interval divided by the total beat length. This calculation integrates the event interval to measure the matching accuracy.

[0069] The rhythmic trajectory index is obtained through rule mapping. For example, if the occupancy continuity of adjacent slots and the overlap of occupancy with the beat are both higher than the preset continuity threshold (the continuity threshold is determined by statistical analysis of historical interference patterns, taking the minimum value that can cover 80% stable overlap), then the rhythmic trajectory index is mapped to the preset high value end; otherwise, the mapping is gradually reduced to ensure that the index range is a dimensionless output of zero to one, which is convenient for subsequent learner input.

[0070] 3-2) Generate the static and stable duration.

[0071] Based on the rhythmic trajectory index and the distribution of stable segments in the preceding module, a stable connectivity interval spanning the handover window is located for each candidate frequency band. This interval is composed of consecutive stable segments and is long enough to accommodate the handover operation. Next, the shortest acceptance duration of this stable connectivity interval is taken as the stable duration. The longer this duration, the more reliable the acceptance of the stable state, making it easier to determine the holding duration. This location and value-taking process focuses on the cross-window connectivity of segments, ensuring that the decision considers the buffer requirements of the actual handover and avoiding a re-handover caused by a brief improvement.

[0072] For each frequency band in the candidate frequency band set, firstly, a set of statically stable segments is selected, where the length and start time stamp of each statically stable segment are recorded; the statically stable connected interval is located by accumulating the total length of adjacent statically stable segments and checking whether it exceeds the switching window threshold (the switching window threshold is set, for example, through the service short message transmission delay test, and is taken as 50% of the end-to-end delay tolerance limit). If it is satisfied, a statically stable connected interval is formed, including the specific range from the start statically stable segment to the end statically stable segment;

[0073] Then, scan all possible sub-intervals within the statically stable connected interval, calculate the duration of each sub-interval (the difference between the end time marker and the start time marker), and select the minimum value as the statically stable duration. This minimum value selection highlights the most conservative robust evaluation, ensuring priority retention at the edge of interference. The statically stable duration is output in milliseconds as the second quantity of the ordinal input, paired with the rhythmic trajectory index, which facilitates the subsequent segmentation of the segmented tree.

[0074] 3-3) The ordinal input segmented tree learner generates switching holding coefficients.

[0075] The generated rhythmic convergence trajectory index and static steady-state duration are input into a sequence-based segmented tree learner in a "first judge timing, then judge acceptance" order. This learner uses a tree structure to split nodes, dividing them based on the rhythmic convergence region (high value range) and free region (low value range) of the rhythmic convergence trajectory index, and the sufficient acceptance region (long duration range) and insufficient acceptance region (short duration range) of the static steady-state duration, along with the consistency of their changing directions. Leaf nodes output confidence prototypes (probability values ​​for switching or staying tendencies), which are calibrated and mapped to generate switching holding coefficients. These coefficients range from -1 to +1, with positive values ​​indicating switching tendencies and negative values ​​indicating staying tendencies. This sequence-based input and splitting mechanism simulates the hierarchical logic of decision-making, ensuring the integrated evaluation of timing and acceptance, and achieving compressed output for short-window discrimination.

[0076] The sequence segmentation tree learner first uses the rhythm-matching trajectory index received by the root node as the basis for the first-level split. If the rhythm-matching trajectory index falls into the rhythm-matching region (e.g., above the 70% threshold, obtained by fitting a pre-trained interference dataset), the right branch is a matching path; otherwise, the left branch is a free path. The second level splits based on the static steady-state duration. If the static steady-state duration falls into the sufficient carrying capacity region (e.g., exceeding a multiple of the switching window threshold, determined by simulating switching success rate statistics), the further branch is a sufficient path; otherwise, it is an insufficient path. The splitting process also checks the consistency of the change direction. When the rhythmic trajectory index increases and the static steady-state duration lengthens, the consistency marker (value is the logarithm of the consistency direction divided by the logarithm of the total change) is enhanced. This check integrates dynamic trends. After traversing the tree to the leaf node, the confidence prototype (switching tendency probability and residence tendency probability, which are complementary) is output. Then, through calibration mapping, the switching tendency probability is multiplied by two and then subtracted by one to obtain the switching holding coefficient. This mapping converts the probability into a coefficient, ensuring positive and negative symmetry, which facilitates threshold judgment. The entire learner uses the preceding training data to initialize the split threshold, ensuring generalization stability under short-term fluctuations in the power line medium.

[0077] The ordinal segmented tree learner is a variant of the decision tree specifically designed for ordinal input. It processes the input through a hierarchical splitting mechanism, outputting confidence in switching or staying tendencies and mapping them to decision coefficients. The following sections elaborate on its construction, training, optimization, and parameter settings.

[0078] Construction Process: The ordinal segmented tree learner is constructed by first defining a tree structure, including a root node, internal nodes, and leaf nodes. The root node receives the first ordinal input (e.g., rhythmic convergence trajectory index), and subsequent nodes process the second input (e.g., static stability duration) in ordinal order. Splitting rules are based on predefined region partitioning: for the first input, its value range is divided into a rhythmic convergence region (representing a high-value interval where disturbance and rhythmic stability overlap) and a free region (representing a low-value interval where disturbance is unstable); for the second input, it is divided into a sufficient-load region (representing a long-duration interval where static stability is reliable) and an insufficient-load region (representing a short-duration interval where static stability is insufficient). Each internal node checks the consistency of the change directions of the two inputs during splitting, such as whether the upward or downward trends of the input values ​​are synchronized, to enhance the robustness of the branch. The tree depth is typically limited to three to five levels to avoid overfitting. Leaf nodes store confidence prototypes, i.e., the probability distribution of switching or staying tendencies. During construction, the splitting threshold needs to be initialized. For example, the threshold of the first input should be set to the middle to high point of the exponential range to ensure that the congruent region covers the majority of stable overlapping samples.

[0079] For example, in power communication, assuming the input is a rhythmic convergence trajectory index (range 0 to 1) and a static steady-state duration (in milliseconds), when constructing the tree, the root node splits into convergent paths (right branch) and detached paths (left branch) based on whether the index is greater than 0.7. Second-level nodes split into sufficient paths and insufficient paths based on whether the duration exceeds 1.5 times the switching window threshold. Furthermore, if the index increases and the duration lengthens, the branch probability is weighted during consistency checks to improve sensitivity to short-term disturbances. This construction ensures that the tree logic simulates a decision-making process of "first determining the timing, then determining the acceptance."

[0080] Training Process: The ordinal segmented tree learner is trained using a supervised learning method with a labeled dataset. Each sample contains an ordinal input pair (rhythm-consistent trajectory index and static steady-state duration) and a target label (switching tendency or residence tendency). Training begins at the root node, recursively splitting the sample set: calculating the purity gain for each potential split point, using metrics such as information gain or the Gini index, but prioritizing ordinal constraints to ensure the first input dominates the initial split. Consistency of change direction is incorporated into the gain formula by calculating the differential trend of the input sequences; for example, if the trends are synchronized, the priority of the split is increased. Training iterates until the purity of the leaf node samples reaches a threshold or the tree depth limit is reached. The confidence prototype is calculated using the label ratio of the leaf node samples; for example, the switching tendency probability is the number of samples of that class divided by the total number of samples. The dataset can be generated by simulating power line interference environments, including historical observations of multi-band short-window state sequences.

[0081] For example, in the training example, suppose the dataset contains 1000 samples, each labeled as "switching" or "staying". After the root node splits, the gain calculation of the confluence subset shows that if 80% of the samples with an exponent greater than 0.7 are switching-prone, this threshold is prioritized. In the second split, if 95% of the subsets with a time interval greater than 150 milliseconds and consistent with the exponent trend are switching-prone, that branch is fixed. After training, the leaf nodes output confidence prototypes, such as a leaf node probability of 0.85 for switching and 0.15 for staying. This process ensures that the learner adapts to the short-term fluctuations of the power distribution scenario, achieving high-accuracy tendency prediction.

[0082] Optimization Process: Optimizing the ordinal segmented tree learner is primarily achieved through post-pruning and cross-validation to reduce overfitting and improve generalization ability. Post-pruning starts from the complete tree and evaluates the performance gain of replacing subtrees with leaf nodes layer by layer, using the accuracy of the validation set or the loss function as an indicator. For example, if the overall accuracy improves after replacement, pruning is performed. Simultaneously, the weights for consistency in the direction of change are optimized, adjusting their contribution to the splitting gain through grid search to ensure priority for stabilizing trends in environments with fluctuating interference. Cross-validation uses a k-fold method (e.g., k=5), alternating between training and validation using portions of the dataset, selecting the tree configuration with the best average performance. Furthermore, an early stopping mechanism can be introduced; if the validation loss no longer decreases in consecutive iterations, the optimization is terminated.

[0083] For example, in the optimization example, the initial tree depth was five levels. After post-pruning evaluation, replacing a subtree containing free and insufficient regions with leaf nodes improved the validation set accuracy from 85% to 92%, while the consistency weight was adjusted from 0.5 to 0.7, further reducing sensitivity to noisy samples. This optimization reduces invalid handovers in power communication applications, ensuring robustness of decisions under short message density transmission.

[0084] Parameter settings include split threshold, tree depth, purity threshold, and consistency weight. The split threshold is determined based on statistical data distribution. For example, the threshold for the first input is determined by histogram analysis, taking the higher end of the median (e.g., 0.7). The threshold for the second input is determined by referencing a multiple of the switching window duration (e.g., 1.5 times, set through business latency testing). The tree depth is set to three to five to balance complexity and computational efficiency. The purity threshold is set to 0.95, indicating that splitting stops when the purity of the leaf node labels reaches this value. The consistency weight is initially 0.5 and adjusted through sensitivity analysis. For example, if increasing the weight makes the model respond more accurately to trend changes, this value is fixed. All parameters can be tuned on the validation set using grid search or Bayesian optimization to ensure optimal performance in specific scenarios.

[0085] For example, in the parameter setting example, for the power line medium scenario, the first input threshold is set to 0.7 (covering 70% of stable samples), the second input threshold is set to 200 milliseconds (exceeding the average switching window), the consistency weight is 0.6 (verified by grid search to improve accuracy by 5%), and the tree depth is four. This setting ensures the learner's parameter adaptability under multi-band interference, avoiding decision lag or jitter.

[0086] 3-4) Perform a switch or keep-alive.

[0087] Based on the generated handover holding coefficients, for each candidate frequency band, if the handover holding coefficient indicates a handover tendency (e.g., the handover holding coefficient is greater than zero), the frequency band with the highest handover holding coefficient is selected from the candidate frequency band set as the target frequency band and a handover operation is performed; if it indicates a retention tendency (the handover holding coefficient is less than zero), the state of the current frequency band indicated by the retention flag is maintained. This execution logic directly responds to the coefficient output, ensuring the immediate implementation of decisions and avoiding a disconnect between evaluation and action.

[0088] The system iterates through all handover holding coefficient values ​​in the candidate frequency band set, selects the frequency band corresponding to the largest positive value as the target frequency band, and triggers handover instructions, including frequency band reconfiguration and data migration. This selection prioritizes high-confidence handover to capture opportunities for improvement. In cases where negative values ​​dominate, the system reads the static stability connectivity of the retention flag. If the static stability connectivity is still higher than the static stability threshold, the current frequency band is locked and the holding duration is updated. This maintenance operation buffers short-term disturbances. After execution, a decision log is recorded, including the handover holding coefficient value and action type, which is convenient for subsequent verification modules to reference, ensuring continuity in scenarios with dense short-message traffic.

[0089] At the merging and discrimination point, two quantities are generated for each candidate frequency band and fed into the sequence segment tree learner to obtain the handover holding coefficient. Quantity 1 is the rhythmic convergence trajectory index. First, the event scale is read and projected onto the detection beat slot to form an occupancy map. Then, the rule mapping is completed by the occupancy continuity of adjacent slots and the overlap tightness of occupancy with the beat. The closer the index is to the convergence end, the more stable the interference and the beat are, and the more predictable it is. Quantity 2 is the static stable duration. First, the static stable connected interval that can cross the handover window is located based on the distribution of static stable segments. Then, the duration of the connection is taken. The shortest acceptance time in the communication interval is used as the time interval. The longer the time interval, the more stable the acceptance. The two quantities are entered into the sequence segmentation tree learner in the order of "first judge the timing and then judge the acceptance". The splitting is based on the consistency of rhythm matching area and free area, sufficient acceptance area and insufficient area and their change direction. The leaf node outputs the confidence prototype of "switching tendency" or "staying tendency", which is calibrated and mapped to the switching holding coefficient. When the coefficient points to the switching tendency, the target frequency band is selected from the candidate frequency band set and the switching is performed. When the coefficient points to the staying tendency, the state pointed to by the staying mark is maintained.

[0090] Through the above steps, the sequence position fusion module realizes the sequence position fusion and coefficient generation of the short window discrimination results, compresses the information of interference rhythm and static stability, ensures the timely and robust execution of the output drive of the switching holding coefficient, significantly reduces back-switching jitter and latency spikes, and ensures the continuous transmission of control messages and service data.

[0091] The switching holding coefficients are split and mapped by the ordinal segmented tree learner to accurately point to the switching or holding tendency, and drive the selection and execution of the target frequency band, thereby realizing the compressed output of the short window discrimination result and the synchronization of the action. The table write-back module takes the switching action triggered by this coefficient as the starting point, collects the before and after frequency band sequences within the switching window and establishes the table relationship. Through consistency verification and review correction, the data caliber is unified to ensure the immediate correction of the execution deviation, and provides a reliable sequence basis after write-back for the iterative update and fine-tuning exploration of the strategy convergence module.

[0092] The specific processing logic of the table write-back module:

[0093] 4-1) Open the switching window and collect the sequence synchronously.

[0094] After the handover holding coefficient generated by the sequence-based decision-making module indicates the handover tendency and the target frequency band is selected, the handover action is immediately triggered, and a handover window is opened simultaneously. This window is defined as a fixed-duration observation period used to isolate the interference effects of the handover process. The acknowledgment sequence (continuous acknowledgment time value sequence) and error frame indication sequence (continuous error frame indication bit sequence) of the frequency band before handover are simultaneously acquired, as well as the corresponding sequence of the target frequency band. These sequences are acquired based on the detection beat interval of the short-window weaving module, ensuring that the acquisition points are aligned with a unified time scale. The synchronous acquisition mechanism captures the instantaneous state changes during handover, facilitating subsequent alignment and avoiding the accumulation of time misalignments.

[0095] The switching window opens in milliseconds, for example, with a length set to twice the expected switching operation duration, and the starting time stamp is locked by a hardware timer. For the frequency band before switching, a receipt sequence is collected, including the time from the first receipt to the last receipt (the number of probe points in the window represents the total number of points), and a frame error prompt sequence, including the time from the first frame error prompt to the last frame error prompt, with each receipt time and frame error prompt corresponding to a time stamp point. Similarly, for the target frequency band, the corresponding receipt sequence and frame error prompt sequence are collected. The acquisition process uses a parallel buffer queue to store the raw data. If the queue is full, the earliest point is overwritten, thus ensuring data integrity in high-interference environments and providing a dual-frequency band timing basis for establishing table relationships.

[0096] 4-2) Establish table relationships using a unified time anchor.

[0097] Based on the acknowledgment sequence and the error frame notification sequence, a unified time anchor is used to align three aspects: message arrival order alignment (matching the message order before and after the handover), acknowledgment time landing point alignment (time stamp synchronization of the acknowledgment time point), and error frame notification edge alignment (matching the edge position of the error frame notification transition point). The table relationship consists of these three aspects, forming a structured alignment record used to quantify handover consistency. This alignment process corrects for timing deviations under media interference, ensuring accurate comparison of the states of the preceding and following frequency bands.

[0098] First, message arrival order alignment is performed. For the acknowledgment sequence of the pre-switching frequency band and the acknowledgment sequence of the target frequency band, the order matching degree is calculated. This value is the number of matching message pairs divided by the total number of messages. The number of matching message pairs is the number of acknowledgment time points with consistent order and their corresponding acknowledgment time points. This calculation emphasizes the continuity of transmission order. Second, acknowledgment time landing point alignment is performed. This is done by projecting the acknowledgment time of the pre-switching frequency band onto a unified time anchor. If the absolute value of the difference between the acknowledgment time and the corresponding acknowledgment time is less than the tolerance (the tolerance originates from a subset of the clock tolerance band, determined through link delay distribution testing, and the average value is taken), then the acknowledgment time is aligned. If the deviation is 20%, it is marked as an aligned landing point, forming a landing point alignment set, which includes multiple aligned landing points; finally, for frame error prompt edge alignment, the rising edge positions of the frame error prompt sequence of the frequency band before the switch and the frame error prompt sequence of the target frequency band are scanned. If the time scale difference between the edges is less than the edge threshold (the edge threshold is set by frame error event statistical analysis, taking 30% of the average short window interval), it is registered as an aligned edge, including multiple edge alignments; the table relationship includes the order matching degree, landing point alignment set and edge alignment, and the integrated records provide a multi-dimensional benchmark for verification, avoiding the deviation of a single alignment.

[0099] 4-3) Calculate the consistency and verify the switching results.

[0100] Using a table-based relationship, the consistency of the three is calculated. This consistency is quantified by fusing the matching ratio of message arrival order alignment, the coverage of acknowledgment time landing point alignment, and the edge overlap rate of error frame prompt edge alignment, and serves as the basis for verifying the handover result. If the consistency is lower than the fallback threshold, it is determined to be in a fallback state, triggering correction.

[0101] Consistency is calculated as follows: the matching ratio of message arrival order alignment, the coverage of acknowledgment time landing point alignment, and the edge overlap rate of error frame notification edge alignment. Each of these is then multiplied by its corresponding weight. The matching ratio of message arrival order alignment is a dimensionless value ranging from zero to one. The coverage of acknowledgment time landing point alignment is the dimensionless value of the number of aligned landing points divided by the total number of landing points. The edge overlap rate of error frame notification edge alignment is the dimensionless value of the number of aligned edges divided by the total number of edges. This averaging ensures a balanced contribution from all three factors, with all physical dimensions being dimensionless, facilitating threshold comparison. The fallback threshold is determined, for example, through statistical experiments simulating fallback scenarios, taking the critical point where a decrease in consistency leads to latency exceeding business tolerance. The verification process iterates through each component in the table relationship. If the consistency is less than the fallback threshold, a fallback flag is marked as one, and the fallback timestamp is recorded. This determination directly links to the error correction initiation, ensuring rapid response to deviations in short message transmission.

[0102] 4-4) Initiate review and correction and register the revised rules.

[0103] When the verification shows a decline (the decline flag is one), the review and correction are initiated. The detection beat slots and event scales are examined in reverse order along the switching window. First, the phase misalignment between the beat and the event (the offset between the detection beat slot position and the event scale sequence) is detected, and a beat correction rule is recorded. Second, the offset of the table landing point is detected (the deviation between the landing point in the alignment set and the unified time anchor exceeds the tolerance), and a landing point correction rule is recorded. These rules record the offset magnitude and direction for subsequent data unification.

[0104] The process reverses time from the end time marker of the window back to the beginning time marker, comparing the slot index in the detection clock slot with the order of the event scale point by point. If the absolute value of the difference between the slot index and the order is greater than the phase threshold (the phase threshold is set, for example, through event interval distribution analysis, taking 40% of the average event interval), then a clock correction rule is registered, including misaligned slots, correction direction (positive or negative), and magnitude (slot difference). This registration highlights the restoration of phase synchronization. At the same time, for each landing point in the landing point alignment set, if the landing point deviation is greater than the tolerance, then a landing point correction rule is registered, including offset landing point, correction time marker, and offset amount. This detection focuses on landing point accuracy. The correction process generates a rule set, including clock correction rules and landing point correction rules, and temporarily stores them in a buffer. This reverse inspection mechanism captures the cumulative error at the moment of switching, ensuring the comprehensiveness of the rules.

[0105] 4-5) Write back the correction results to the frequency band short window state sequence.

[0106] The rule set and the short-window trajectories observed within the switching window (window subsequences derived from the receipt time trajectory and error frame indication bits) are written back to the frequency band short-window state sequence to unify data caliber and ensure consistent input for subsequent modules. The write-back operation integrates real-time observation and correction rules, facilitating iterative updates of the policy convergence module.

[0107] Extract the short window trajectory within the switching window, including the window receipt time trajectory (a subset of the receipt time trajectory within the window) and the window error frame indication bits (a subset of the error frame indication bits); then, for the corresponding short window in the frequency band short window state sequence, insert a rule set, such as adding a beat correction rule after the detection beat slot and a landing point correction rule after the event scale, and mark the write-back time stamp next to the short window trajectory; the write-back uses an atomic update operation, and if there is a conflict, the rule is overwritten first, thereby ensuring the version consistency of the frequency band short window state sequence; after the write-back is completed, update the sequence metadata, including the write-back count and the last correction time stamp.

[0108] When a handover action is triggered, the handover window is opened, and the acknowledgment sequence and error frame prompt sequence of the frequency band before handover and the target frequency band are collected synchronously. A table relationship is established with a unified time anchor. The table relationship consists of message arrival sequence alignment, acknowledgment time landing point alignment, and error frame prompt edge alignment. The consistency of the three is used as the basis for handover result verification. When the verification shows a drop, the review correction is started. The clock slot and event scale are checked in reverse order along the handover window. When the clock and event phase misalignment is found, the clock correction rule is registered. When the table landing point offset is found, the landing point correction rule is registered. The above rules and the short window trajectory observed in the handover window are written back to the frequency band short window status sequence to unify the data caliber.

[0109] Through the above steps, the table write-back module realizes the establishment of table relationships and real-time write-back correction within the switching window, unifies the data caliber of the front and back frequency bands, ensures accurate correction of short window trajectories, significantly reduces retransmission accumulation and latency spikes caused by execution deviations, and enhances the intuitiveness of business continuity and maintenance positioning.

[0110] The frequency band short-window state sequence is verified and corrected by a unified time anchor, accurately integrating the short-window trajectory and correction rules within the switching window, thus providing a long-term observation basis with consistent data caliber. The strategy convergence module uses this sequence as input, iteratively revising the candidate patterns and retention markers during idle periods, and verifying the field consistency of the beat and landing point correction rules through small-volume micro-tuning. Finally, a stable switching strategy and time series execution record are generated, ensuring that the strategy dynamically converges according to the interference behavior of the power line medium, and providing closed-loop support for the long-term adaptation and maintenance traceability of the entire multi-band transmission system.

[0111] The specific processing logic of the strategy convergence module:

[0112] 5-1) Read the frequency band short window state sequence after writing back.

[0113] After the table write-back module completes the write-back, it enters an idle period (defined as a low-traffic period with no service load, such as a period when the control message interval exceeds the preset idle threshold). It reads the frequency band short window state sequence corrected by the write-back, which contains the inserted rule set (including the beat correction rule and the landing point correction rule) and the short window trajectory.

[0114] Idle period detection monitors service traffic. If the traffic is below the idle threshold (determined, for example, by historical load statistics, taking 80% of the average packet interval), the read buffer is activated. The subsequence after the last write-back time stamp is located from the metadata of the frequency band short window state sequence. This includes multiple update short windows, each of which integrates the short window trajectory, beat correction rules, and landing point correction rules. The read process uses a sequence snapshot mechanism to copy the subsequence to a temporary storage area and simultaneously verify the sequence integrity. If any is missing, the process rolls back to the previous write-back time stamp, ensuring the atomicity of data in a multi-threaded environment and providing a reliable basis for subsequent revisions.

[0115] 5-2) Iteratively update the candidate patterns and retention markers.

[0116] Based on subsequences, candidate patterns (derived from the static and stable connectivity patterns and perturbation rhythms of the stable plotting module, i.e., the interval distribution of event scales) and retention markers are iteratively updated, revising the inclusion boundaries: candidate patterns have their static and stable boundaries revised according to their static and stable connectivity patterns (adjusting the lower limit of static and stable connectivity) and perturbation boundaries revised according to their perturbation rhythms (adjusting the periodic threshold of event scale intervals); retention markers have their boundary attention (attention interval for consistency) and retention time (upper limit of holding duration) revised according to the switching window verification statistics. This iteration integrates write-back correction, improving the on-site adaptability of patterns.

[0117] Initialize the iteration count to zero. For candidate patterns, first extract the set of stable segments from the subsequence and calculate the revised stable boundary. Its value is the current stable connectivity multiplied by one minus the updated value of perturbation and hybridity. The current stable connectivity is a dimensionless value, and the updated value of perturbation and hybridity is the number of perturbation segments divided by the total number of segments (dimensionless). This multiplication operation integrates the relative proportions of connectivity and hybridity to ensure that the revised stable boundary is a dimensionless boundary value. For perturbation rhythms, scan the interval sequence of the event scale and revise the perturbation boundary. Its value is the average interval divided by the maximum interval. This ratio highlights the periodic stability. For dwell markers, statistically analyze the distribution of consistency in the subsequence and revise the boundary of interest. Its value is the minimum consistency plus the maximum consistency minus the minimum consistency divided by two, and the median is used to adjust the interval. The hold time limit is the current holding time multiplied by one plus the sum of the offsets of the landing point correction rules divided by the switching window length. The sum of offsets is the sum of the deviations of all landing point correction rules in milliseconds, and the switching window length is in milliseconds. This adjustment reflects the correction effect.

[0118] The iteration loop continues until the iteration count reaches the upper limit (e.g., ten times) or the change is less than the convergence threshold (the convergence threshold is set, for example, by simulating the error convergence curve of a long-term mismatch scenario, taking a point where the rate of change is less than 5%). The output revision rules include revising the statically stable boundary and revising the perturbation boundary, as well as revising the dwell flag, including revising the boundary attention and revising the dwell time limit. This iterative revision ensures the gradual optimization of the boundary.

[0119] 5-3) Initiate micro-debugging to verify the revised rules.

[0120] By leveraging revision patterns and revision retention markers, fine-tuning probes are triggered with low-volume traffic to verify the effectiveness of recently revised beat correction and landing point correction rules. Low-volume traffic is defined as a low-intensity, short probe frame sequence (e.g., traffic rate below 10% of normal business). When feedback is consistent, it is solidified into the policy record; when inconsistent, it is backfilled into the rule revision pool. This verification process simulates real-world behavior, corrects long-term mismatches, and ensures the practicality of the rules.

[0121] Fine-tuning probes are initiated during idle periods, generating a small-flow sequence consisting of multiple probe frames (the sequence length is, for example, five times the probe beat interval). The static stability boundary and disturbance boundary are revised according to the revision rules, applying beat correction rules and landing point correction rules. For example, the magnitude of the beat correction rule is offset at the probe beat slot, and the offset of the landing point correction rule is adjusted on the acknowledgment time trajectory. The small-flow sequence is sent to the target frequency band, and feedback sequences are collected, including acknowledgment times and error frame alerts. Consistency is verified by comparing the feedback sequence with the expected trajectory (simulated based on the revision rules), and a... The consistency rate, a dimensionless value, is calculated by dividing the number of matching points by the total number of points. If the consistency rate is greater than the verification threshold (the verification threshold is determined by statistics on the success rate of small-volume tests, taking the lower limit of the 90% confidence interval), it is marked as consistent. When the feedback is consistent, the beat correction rules and landing point correction rules are solidified into the strategy record, including solidified beat correction rules and solidified landing point correction rules. When the feedback is inconsistent, it is backfilled into the rule revision pool, including inconsistent rules and feedback deviation descriptions. The rules in the pool are sorted by priority (decreasing order of deviation magnitude). This probe mechanism captures the actual effect of the revision and avoids the disconnect between theory and practice.

[0122] 5-4) Establish a stable switching strategy and execution record.

[0123] Based on the verification results, the items to be merged in the merge rule revision pool are ultimately formed into a stable switching strategy (integrating revision patterns, revision retention markers, and solidified rules) and execution records. The switching time, target frequency band, verification results, and correction rule number are saved in chronological order for easy traceability. The strategy is solidified through the merging operation to ensure long-term adaptability.

[0124] The rule revision pool is scanned. If the pool is not empty, high-priority rules are selected and merged with the policy records. For example, the deviation description of the inconsistent beat correction rule is injected into the adjustment factor of the revised static stability boundary to generate the final stable switching policy, including the revision boundary set, the solidified rule set, and the retention period. The execution record is recorded as a time-series log, including multiple entries. Each entry contains the switching time, target frequency band, verification result, and correction rule number, where the correction rule number is a unique rule identifier (e.g., sequence number plus version). The log is saved to persistent storage in append mode. If the log exceeds the capacity, the old entries are archived. This serialization ensures the integrity of traceability. After the formation is completed, the update module status is "convergence complete" and the stable switching policy is output for loading into the global policy library. The final step is to lock the on-site behavior matching of the policy.

[0125] During idle periods, the frequency band short-window state sequence, which has been corrected by writing back, is read. Candidate patterns and retention markers are iteratively updated. Candidate patterns are revised and incorporated into the boundary according to the static stable connectivity and disturbance rhythm. Retention markers are revised according to the switching window verification statistics and the boundary attention and retention time limit. The newly revised beat correction rules and landing point correction rules are verified by triggering micro-debugging with small flow. When the feedback is consistent, it is solidified into the strategy record. When it is inconsistent, it is backfilled into the rule revision pool for merging. Finally, a stable switching strategy and execution record are formed. The switching time, target frequency band, verification result and correction rule number are saved in chronological order for easy traceability and review.

[0126] Through the above steps, the strategy convergence module realizes the iterative revision of the write-back sequence and the dynamic solidification of rule verification, corrects long-term mismatches, ensures the on-site convergence of stable switching strategies and the convenience of traceability of execution records, significantly improves the long-term robustness of frequency band selection and business availability, and resolves the lack of granularity in maintenance and reproduction scenarios.

[0127] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0128] It should be noted that the system of the present invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting a variety of hardware environments and usage requirements.

[0129] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0130] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely to distinguish one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart switching and transmission system for multi-band signals in power communication, characterized in that, include: Short window sequencing module: Under a unified time scale, short probe frames are sent sequentially to each frequency band of the same power line medium and the receipt time and error frame prompts are read. The frequency band short window status sequence is generated in chronological order. The continuity and stability plotting module identifies fluctuation boundaries and forms continuity indicators on the short-window state sequence of frequency bands, and outputs the candidate frequency band set and the retention mark of the current frequency band, which together serve as the input for the merging decision; Sequence merging module: At the merge decision point, a handover holding coefficient is generated based on the short window discrimination result. When the coefficient points to a handover tendency, the target frequency band is selected from the candidate frequency bands and the handover is performed. When the coefficient points to a retention tendency, the retention is maintained. Table writing-back module: Collects the table relationship and short window trajectory of the frequency band before and after the handover in the handover window, completes the handover result verification based on the table relationship, performs review and correction when the verification display falls back, and writes the correction result back to the frequency band short window status sequence; Strategy convergence module: Based on the short-window state sequence of the frequency band after write-back, it iteratively updates the candidate patterns and retention markers, initiates fine-tuning probes during idle periods to correct long-term mismatches, and generates stable handover strategies and execution records.

2. The intelligent switching and transmission system for multi-band power communication signals according to claim 1, characterized in that: The short window weaving module performs time anchor alignment on the acknowledgment landing points of different frequency bands. When the time interval of adjacent detections deviates from the beat tolerance band, a beat offset mark is added. The rising edge of the frame error prompt and the spike of the acknowledgment time are registered as event scales. The order and interval of occurrence are recorded with the original time stamp, forming a structured record of short window-slot-event.

3. The intelligent switching and transmission system for multi-band power communication signals according to claim 2, characterized in that: The continuous stabilization plotting module scans the error frame prompt bit and the acknowledgment time trajectory along the time axis on the short window state sequence of the frequency band. When the error frame prompt changes from closed to active or the acknowledgment time exceeds the convergence band, the fluctuation boundary is calibrated, and the static stable segment and the disturbance segment are divided according to the boundary. The continuity indication is calculated based on the segment splicing relationship.

4. The intelligent switching and transmission system for multi-band power communication signals according to claim 3, characterized in that: The continuity and stability plotting module selects frequency bands that meet the static stability requirements to form a candidate frequency band set based on the continuity indication, and writes them into the retention flag according to the static stability connectivity of the current carrying frequency band. The candidate frequency band set and the retention flag are used as inputs for subsequent merging decisions.

5. The intelligent switching and transmission system for multi-band power communication signals according to claim 4, characterized in that: At the merging decision point, the sequence position judgment module generates a rhythmic sync trajectory index and a static stable duration for each candidate frequency band. The rhythmic sync trajectory index first reads the event scale and projects it onto the detection beat slot to form an occupancy map. Then, it completes the rule mapping based on the occupancy continuity of adjacent slots and the overlap of occupancy with the beat. The static stable duration first locates the static stable connected interval that can cross the switching window based on the distribution of static stable segments, and then takes the shortest acceptance time of the corresponding connected interval as the duration.

6. The intelligent switching and transmission system for multi-band signals in power communication according to claim 5, characterized in that: The sequence-based decision module inputs the rhythm-matching trajectory index and the static steady-state duration into the sequence-based segmented tree learner according to the order of timing and acceptance. The splitting criteria are the consistency of rhythm-matching regions and free regions, sufficient acceptance regions and insufficient acceptance regions, and their changing directions. The leaf node outputs of the sequence-based segmented tree learner are confidence prototypes of switching tendency or retention tendency, which are calibrated and mapped to switching holding coefficients. When pointing to switching tendency, the target frequency band is selected from the candidate frequency band set and switching is performed. When pointing to retention tendency, the state pointed to by the retention flag is maintained.

7. The intelligent switching and transmission system for multi-band power communication signals according to claim 6, characterized in that: When the switching action is triggered, the table write-back module opens the switching window and synchronously collects the acknowledgment sequence and error frame prompt sequence of the frequency band before switching and the target frequency band. The table relationship is established with a unified time anchor. The table relationship consists of the alignment of message arrival order, the alignment of acknowledgment time landing point, and the alignment of error frame prompt edge. The consistency of the three is used as the basis for verifying the switching result.

8. The intelligent switching and transmission system for multi-band power communication signals according to claim 7, characterized in that: When the table write-back module shows a drop, it initiates a review and correction. It checks the detection beat slot and event scale in reverse order along the switching window. When a phase misalignment between the beat and the event is found, a beat correction rule is registered. When a table landing point offset is found, a landing point correction rule is registered. The beat correction rule, the landing point correction rule, and the short window trajectory observed in the switching window are all written back to the frequency band short window state sequence to unify the data caliber.

9. A power communication multi-band signal intelligent switching transmission system according to claim 8, characterized in that: During idle periods, the strategy convergence module reads the frequency band short-window state sequence that has been corrected by write-back, and iteratively updates the candidate patterns and retention markers. The candidate patterns are revised and incorporated into the boundary according to the static stable connectivity and disturbance rhythm. The retention markers are revised according to the switching window verification statistics and the attention and retention time limit of the boundary. The newly revised beat correction rules and landing point correction rules are verified by triggering micro-debugging with small flow. When the feedback is consistent, it is fixed into the strategy record. When it is inconsistent, it is backfilled into the rule revision pool for merging. Finally, a stable switching strategy and execution record are formed, and the switching time, target frequency band, verification result and correction rule number are saved in chronological order.

Images