A method and system for precise control of negative pressure in different zones of a ship's suction sail

Abstract

This invention discloses a method and system for precise control of negative pressure in different zones of a ship's suction sail. It quantifies the degree of pulsation suppression in each zone by collecting differential pressure data from each suction zone, identifies abnormal adsorption states and their spread diffusion paths, and establishes a risk map with risk levels. The system then fits the given negative pressure values ​​for each zone using the risk level as a weight, and adaptively calibrates the upper limit of the negative pressure in each zone based on actual overload events of the drive unit. A dynamic calibration map is formed after verification using historical overload distribution. Based on the dynamic calibration map, it implements separate shielding differential pressure coordination and graded under-suction boundary approximation to obtain an attachment flow distribution map characterizing the critical suction force of each zone. Finally, it tracks the contraction of the attachment zone to dynamically identify the lift margin attenuation region, and after verification by global lift constraints, actively reduces speed to save energy and outputs an electronically controlled negative pressure adjustment command, achieving synergy between precise suction distribution and active control of drive power consumption.

Description

Technical Field

[0001] This invention relates to the field of ship energy-saving device control technology, and in particular to a method and system for precise control of negative pressure in different zones of a ship's suction sail. Background Technology

[0002] As a marine energy-saving device that utilizes hydrodynamics to generate propulsive lift, the airflow attachment state at various locations on the suction sail directly affects the overall propulsive efficiency of the sail. In actual navigation, changes in the incoming flow angle, fluctuations in ship speed, and the non-uniform distribution of the sea surface wind field cause differentiated evolution of the flow patterns at different spanwise locations on the sail. The risk of localized airflow separation tends to accumulate after prolonged navigation at a specific angle of attack. Existing control schemes typically apply a uniform suction adjustment strategy to the entire suction area, failing to identify early differences in the airflow attachment state in each area and lacking the ability to anticipate potential separation risks in specific regions. This results in a long-term structural misalignment between suction distribution and the actual needs of each area, leaving areas with abnormal flow patterns on the sail without targeted, proactive protection.

[0003] Existing solutions also have significant shortcomings in terms of drive unit calibration and energy-saving control. Most systems use fixed upper limits for suction force, which are not updated according to the actual load conditions of the drive unit. When mechanical aging or changes in thermal conditions lead to a decrease in load-bearing capacity, the fixed upper limit remains effective, becoming the root cause of long-term overload operation of the drive unit. On the other hand, the sail maintains high suction force even under low lift demand conditions, and the excess suction power cannot be converted into effective propulsion contribution, resulting in low overall system energy efficiency. Summary of the Invention

[0004] This invention discloses a method and system for precise control of zoned negative pressure of a ship's suction sail. The aim is to establish a dynamic mapping relationship between the given suction force of each zone and the actual load limit of the drive unit by sensing and locating the airflow attachment state of each suction zone. Under the premise of maintaining the safety of the wing surface attachment, the power consumption redundancy of each zone is identified and actively reduced, thereby realizing precise zoned negative pressure control of the suction sail and coordinated optimization of system energy consumption under complex navigation conditions.

[0005] The first aspect of this invention proposes a method for precise control of negative pressure in different zones of a ship's suction sail, comprising the following steps: Pressure data is collected independently for each suction zone to form a regional pressure difference sequence. Long-term ultra-stable segments of pressure difference are identified from the regional pressure difference sequence to obtain an abnormal signal sequence of overstability. Based on the statistical analysis of the trigger frequency depth weighting of the overstability anomaly precursor sequence, a separation risk index is generated. The separation risk index is used to screen for long-term pressure stability characteristics to determine hidden separation zones. The hidden separation zones are then used to track the pressure gradient changes between adjacent regions to establish a zonal risk map. Based on the risk map of the partition, a weighted iterative fitting of the sudden change in the angle of attack of the incoming flow is carried out to obtain the negative pressure setpoint of the partition. The negative pressure setpoint of the partition is subjected to overload-triggered high-confidence calibration to construct a self-correcting mapping table. The self-correcting mapping table is verified by historical overload distribution to form a dynamic calibration mapping. Based on the dynamic calibration mapping, a separate shielded full-area differential pressure coordinated regulation output differential pressure regulation scheme is implemented. Based on the differential pressure regulation scheme, a graded under-pumping boundary approximation is performed to obtain the attached flow distribution map. The attachment flow distribution map is used to track the contraction rate of the attachment zone and output a constant speed adjustment set. The lift marginal gain attenuation segment is identified from the speed adjustment set to determine the energy saving target area. Active deceleration and overdrive archiving operations are performed on the energy saving target area to generate energy saving control quantity. The energy saving control quantity is verified by global lift constraint and an electronically controlled negative pressure adjustment command is output.

[0006] A second aspect of this invention provides a precise control system for zoned negative pressure of a ship's suction sail, comprising: The data acquisition module is used to independently acquire pressure data for each suction zone to form a regional pressure difference sequence, and to identify long-term ultra-stable segments of pressure difference from the regional pressure difference sequence to obtain an over-stability anomaly precursor sequence; The risk identification module is used to generate a separation risk index based on the depth weighted average of the trigger frequency of each region according to the overstability anomaly precursor sequence, screen for long-term pressure stability characteristics based on the separation risk index to determine hidden separation areas, and use the hidden separation areas to track the pressure difference gradient changes between adjacent regions to establish a zonal risk map. The calibration mapping module is used to obtain the partition negative pressure given value by performing weighted iterative fitting of sudden change of incoming flow angle based on the partition risk map, implement overload triggered high-confidence calibration of the partition negative pressure given value to construct a self-correcting mapping table, and verify the historical overload distribution of the self-correcting mapping table to form a dynamic calibration mapping. The collaborative allocation module is used to implement a separate shielded full-area differential pressure collaborative adjustment output differential pressure adjustment scheme based on the dynamic calibration mapping, and to perform graded under-pumping boundary approximation based on the differential pressure adjustment scheme to obtain the attached flow distribution map. The negative pressure control module is used to track the contraction rate of the attachment zone and output a constant speed adjustment set through the attachment flow distribution map, identify the lift marginal gain attenuation segment from the speed adjustment set to determine the energy saving target area, perform active deceleration and overdrive archiving operations for the energy saving target area to generate energy saving control quantity, and perform global lift constraint verification on the energy saving control quantity to output an electronically controlled negative pressure adjustment command.

[0007] The beneficial effects of this invention are reflected in the following points: First, by comparing the relative quantification of the differential pressure pulsation in each suction zone with the frequency of historical operating conditions, the long-term ultra-stable areas of differential pressure and their spanwise diffusion paths are incorporated into risk perception, and local attachment anomalies are located to specific spanwise positions during the risk accumulation stage of detachment. Furthermore, using the risk level of each zone as the differential weight for fitting the suction setpoint, the preset suction amount in each zone directly corresponds to its historical flow regime change characteristics, solving the problem of unified adjustment strategies masking early signals and long-term misalignment between suction distribution and actual attachment requirements. Second, the upper limit of suction does not rely on fixed calibration values, but instead uses the actual overload moment as the trigger source to backtrack and extract samples of the actual load-bearing boundary. It is continuously iteratively corrected according to the dual-dimensional operating condition index of angle of attack and ship speed, and verified through historical frequency distribution to achieve tightening in the high-frequency range and release in the low-frequency range. The control boundary maintains dynamic matching with the evolution of load-bearing capacity such as mechanical aging and thermal fluctuations, eliminating the risk of long-term failure of fixed parameters after unit deterioration. Finally, relying on dynamic calibration mapping, separate shielded differential pressure coordinated regulation and graded under-pumping boundary approximation are implemented. Under the premise of maintaining the lower limit of attachment safety, the critical suction distribution is calibrated zone by zone. On this basis, the lift marginal gain saturation is determined zone by zone according to the speed-lift response characteristics. Differentiated deceleration is implemented by combining the credibility level of historical overdrive frequency time reverse decay weighting. After global lift constraint verification, an electronically controlled negative pressure regulation command is output. The redundant pumping power consumption under low lift demand conditions is actively reduced, and the system energy consumption and actual propulsion contribution form a quantifiable closed-loop control. Attached Figure Description

[0008] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0009] Figure 1 This is a flowchart illustrating a method for precise control of negative pressure in a ship's suction sail according to the present invention.

[0010] Figure 2 This is a structural block diagram of a ship suction sail zoned negative pressure precision control system according to the present invention. Detailed Implementation

[0011] 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, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0012] 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.

[0013] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0014] The technical solutions of the embodiments of this application will be described below.

[0015] like Figure 1 As shown, this embodiment of the invention provides a method for precise control of negative pressure in a ship's suction sail, including the following steps S101-S105: Step S101: Independently collect pressure data for each suction zone to form a regional pressure difference sequence, and identify the long-term ultra-stable segment of the pressure difference from the regional pressure difference sequence to obtain the over-stability anomaly precursor sequence.

[0016] Specifically, pressure data is independently collected for each suction zone to form a regional pressure differential sequence. A pair of measuring points are arranged in each suction zone: the static pressure port inside the negative pressure chamber and the static pressure orifice on the outer surface of the airfoil. The pressure data for that zone is output via a differential pressure transmitter and written into a ring-shaped time-series buffer according to the sampling period. The time-series pressure differential data for all suction zones are arranged side-by-side according to their zone numbers to form the regional pressure differential sequence. The difference between adjacent sampled values ​​represents the pulse intensity of that zone, and the pulse amplitude of the regional pressure differential sequence is a direct basis for judging the health status of the adsorption boundary layer in each zone. Under normal upwind operation, the regional pressure differential sequence of each zone exhibits a continuous oscillating pattern, with the pulse intensity fluctuating slightly with the incoming turbulence. Once the sail angle remains stationary at a certain angle of attack for an extended period, the dynamic interaction between the overall incoming flow and the sail surface weakens, and the pulse amplitude of the regional pressure differential sequence in most suction zones will narrow synchronously. This overall low-pulsation pattern is similar to the low-pulsation pattern caused by local flow separation in a single area. The difference lies in that the former shows a synchronous decrease in the absolute amount of pressure difference across all areas and a uniform spatial distribution, while the latter only occurs in one or two adjacent areas and the pulsation in the surrounding areas does not decrease synchronously. A horizontal comparison in the spatial dimension can preliminarily distinguish between the two sources. During the acquisition phase, the pressure data output by the sensors in each area are checked for synchronization consistency: when the reading of a single point deviates from the median of the entire area by more than a set multiple of the standard deviation (e.g., 3 times), the average value of the adjacent time points is used to replace the sampling point. Sensor drift is thus isolated from the regional pressure difference sequence, leaving no false low-pulsation segments. Data gaps caused by single-area sensor failures are marked with gaps and are not included in the calculation of the median benchmark for the entire area. The continuous output of the regional pressure difference sequence for other areas is unaffected.

[0017] In some embodiments, the step of identifying long-term ultra-stable segments of pressure differential from the regional pressure differential sequence to obtain an over-stability anomaly precursor sequence includes: performing insufficient pulsation intensity quantification on the regional pressure differential sequence to obtain a pressure differential stability measure; performing a screening of persistent high-value segments of pressure differential stability to determine long-term ultra-stable segments of pressure differential; conducting historical ultra-stable frequency comparison verification based on the long-term ultra-stable segments of pressure differential to obtain a stable precursor sequence; and calibrating the spread-directional stable anomaly diffusion region based on the stable precursor sequence to form an over-stability anomaly precursor sequence.

[0018] To obtain a measure of pressure stability, a quantification of insufficient pulsation intensity was performed on the regional pressure differential sequence. The standard deviation of each region's pressure differential sequence was calculated window-by-window using a fixed-length sliding window, with a window length of 64 sampling points covering approximately 3 to 4 seconds. The standard deviation fluctuated continuously with the sail attachment state. A consistently low standard deviation indicated that the pulsating components excited by the incoming turbulence in that region were abnormally suppressed. A normally adsorbed boundary layer would not remain in this low-energy static state for an extended period under continuous turbulent disturbance. Looking only at the absolute value of the standard deviation could lead to systematic misjudgment: when sea conditions are stable and the sail angle is fixed for a long period, the overall incoming turbulence intensity decreases, and the standard deviation of the regional pressure differential sequence in most suction zones decreases synchronously. This synchronous sinking across the entire region is inseparable from the narrowing of a single region caused by local flow separation in absolute terms. Therefore, the standard deviation of each region was divided by the median standard deviation of the entire region at the same time to obtain the relative pulsation intensity. Then, 1 was subtracted from the relative pulsation intensity to obtain a measure of pressure stability. A larger measure of pressure stability indicates that the pulsation intensity of that region is lower than the overall benchmark for the entire region; when it approaches 1, it corresponds to the almost complete disappearance of pulsation in that region. In the regional differential pressure sequence, missing frames caused by brief sensor outages are filled with the weighted average of the preceding and following moments. These filled frames are not included in the calculation of the standard deviation for their respective regions or the median for the entire region. Otherwise, smooth filling would lower the standard deviation and the benchmark, causing the differential pressure stability measurement in normal regions to appear artificially high. The length of the period of consecutively high differential pressure stability measurements directly corresponds to the depth at which the adsorption boundary layer in that region deviates from normal oscillations. A normal overall regional mean but a persistently high differential pressure stability measurement in a single region indicates that local separation risks are accumulating in isolation within an overall healthy surface. A synchronous increase in differential pressure stability measurements in two adjacent regions further suggests that the separation region already exhibits early characteristics of longitudinal diffusion.

[0019] The pressure differential stability measurement is used to screen for persistently high values ​​to determine long-term ultra-stable pressure differential segments. Only segments where the pressure differential stability measurement continuously exceeds a threshold and the number of frames reaches a set lower limit are considered long-term ultra-stable pressure differential segments. The lower limit of the frame number can be calculated by sampling 200 frames at 50 milliseconds, approximately 10 seconds. This sustained frame number threshold distinguishes occasional low-pulsation spikes from true flow separation precursor signals. The threshold is determined by referring to the 95th percentile of the pressure differential stability measurement in each region of historical normal navigation data. When the incoming flow is briefly blocked or the sail angle is slightly adjusted, the pressure stability measurement in the disturbed area often spikes and then quickly falls back within a few frames. If such instantaneous spikes are treated as long-term ultra-stable pressure segments, they will repeatedly generate false alarms. The continuous frame threshold is designed to address this situation. The threshold value is based on the shortest occurrence duration of the flow separation precursor signal in history. If it is too short, brief disturbances will frequently trigger candidate segments. If it is too long, the starting boundary of the long-term ultra-stable pressure segment will be identified by hysteresis. Therefore, a small margin is left on the basis of the shortest duration. If there is an intermittent crossing within the candidate segment where the pressure stability measurement briefly falls back below the threshold and then rises again, if the number of frames of the crossing does not exceed the set tolerance (single-digit frame level), it is considered a small fluctuation within the segment, and the candidate segment remains continuous. If it exceeds the tolerance, it is truncated at the crossing point into two independent long-term ultra-stable pressure segments. The design logic for maintaining continuity is as follows: during the actual flow separation precursor period, sensor signals are inevitably subject to brief interference. Forcibly truncating them would break the same continuous risk event into multiple short segments, underestimating the actual duration of a single long-term ultra-stable pressure differential segment. In addition, the leeward edge region itself has lower pulsating energy, and its pressure differential stability measurement threshold is usually higher than that of the windward core region. Applying a uniform threshold to the entire sail would cause the long-term ultra-stable pressure differential segment in the edge region to be systematically over-triggered. Therefore, the threshold is determined independently for each zone.

[0020] Stability precursor sequences are obtained by comparing and verifying historical ultra-stable frequencies based on the long-term ultra-stable period of differential pressure. The ratio of the current long-term ultra-stable period frequency of differential pressure in each region to the historical reference frequency under the same operating conditions constitutes the overfrequency ratio. A consistently high overfrequency ratio indicates that the ultra-stable events in this region under the current operating conditions have deviated from the historical normal range and can no longer be explained by the overall weakening of the inflow. The operating conditions are matched using the inflow angle of attack range and ship speed range as a two-dimensional index, and the operating condition grid can be divided into granularities of 2 degrees angle of attack and 2 knots ship speed. The necessity of differentiating operating conditions lies in the fact that the leeward edge region inherently exhibits weaker downstream adhesion at high angles of attack. Historical data shows that the frequency of long-term ultra-stable pressure differential segments in such regions is naturally higher. Directly labeling the current long-term ultra-stable pressure differential segment as abnormal without differentiating operating conditions would lead to continuous false alarms in the edge region. By referencing the historical frequency benchmark corresponding to the current operating condition, this systematic bias is absorbed by the over-frequency ratio mechanism. Only regions with an over-frequency ratio continuously exceeding a threshold are identified as abnormally ultra-stable regions under this operating condition, and their ultra-stable records are subsequently added to the stability precursor sequence. The historical frequency database is updated continuously with accumulated flight data, using flight cycles as the unit of measurement. When the historical sample size of long-term ultra-stable pressure differential segments in a certain operating condition interval is insufficient to meet the set minimum value, the frequency benchmark is estimated using the weighted average of neighboring operating condition intervals, and the corresponding entries are marked with a low-confidence label in the stability precursor sequence. Ultra-stable records marked with this label are also included in the stable precursor sequence. However, since the estimation benchmark is inevitably biased, these records are weighted down by 0.7 times during the risk index calculation phase to prevent frequency fluctuations in the normal range from being amplified to artificially high risk levels. The overfrequency ratio threshold is gradually tightened based on the actual false alarm rate as historical data accumulates. The version number of the stable precursor sequence is linked to the corresponding threshold record for historical tracing and comparison at different accumulation stages.

[0021] Based on stable precursor sequences, a longitudinally stable anomaly diffusion zone is calibrated to form an overstable anomaly precursor sequence. The core of longitudinal diffusion calibration is the temporal synchronicity of ultrastable events in adjacent anomaly zones within the stable precursor sequence. Isolated ultrastable events in a single zone and progressive diffusion in multiple zones exhibit distinctly different spatial morphologies in their longitudinal distribution. Under normal operating conditions, the flow states in each pumping zone are independent. When one or two zones are ultrastable, the pulsation in the surrounding areas is generally unaffected. If the ultrastable events in two or more adjacent zones in the stable precursor sequence highly overlap in time (the overlap duration accounts for a set value, which can be taken as 70%), and the starting time of the ultrastable segment in each zone shifts monotonically backward along the longitudinal direction, and the starting delay between adjacent zones falls within a set window, it can be determined that low pulsation is spreading longitudinally from the initial region, and flow separation is evolving from single-point failure to multi-zone continuous failure. Once this diffusion path is identified, the entire covered area is classified as a longitudinally stable anomaly diffusion zone, and the corresponding time period, area range, and original labeling are subsequently written into the overstable anomaly precursor sequence. In a stable precursor sequence, if an isolated single-area is extremely stable while adjacent areas exhibit normal synchronous pulsations, it does not trigger the directional diffusion determination. This area is recorded separately in the overstable anomaly precursor sequence with a single-area anomaly marker. If, in a subsequent period, a cascading pattern occurs where one area recovers pulsations first, followed by instability in the remaining areas, the overstable anomaly precursor sequence adds a cascading instability marker to the corresponding period. The appearance of this marker indicates that the overall sail deterioration rate is much faster than isolated single-area failure; once multi-area synchronous low pulsations are triggered, the effective lift of the sail can drop dramatically in a very short time.

[0022] Step S102: Based on the statistical analysis of the trigger frequency depth weighting of the overstability anomaly precursor sequence, a separation risk index is generated. The separation risk index is used to screen for long-term pressure stability characteristics to identify hidden separation zones. The hidden separation zones are then used to track the pressure gradient changes between adjacent regions to establish a zonal risk map.

[0023] Specifically, a separation risk index is generated based on the depth-weighted amount of the trigger frequency of each region according to the statistical analysis of the overstable anomaly precursor sequence. The cumulative count of triggering events in the overstability anomaly precursor sequence of each region within the sliding time window is normalized after being weighted and summed with the corresponding ultrastability duration depth. The separation risk index is calculated according to the following formula: DRI(i,t)=[1 / W_total]×Σw(k)×d(i,k), k∈Ω(i,t), where DRI(i,t) is the separation risk index of region i within the sliding window at time t; Ω(i,t) is the set of all triggering events in region i within the window at time t; w(k) is the weight coefficient of the k-th triggering event, which is 1.5 for cascade instability labeled events, 0.7 for low-confidence labeled events, and 1.0 for ordinary events; d(i,k) is the window mean of the pressure difference stability measurement of region i during the k-th triggering event, reflecting the duration depth of the ultrastability event; W_total is the normalization benchmark, which is the mean of d(i,k) of all triggering events in the current window, and DRI is recorded as zero when there is no event. When a ship maintains a high angle of attack for an extended period in a sustained crosswind, the suction zone at the boundary between the windward and leeward sides of the sail repeatedly experiences a cycle of brief hyperstability followed by recovery. In this zone, Ω(i,t) accumulates continuously over tens of minutes, causing the DRI(i,t) to rise accordingly. Meanwhile, the DRI(i,t) in other areas remains low. This spanwise difference directly points to the localized location of latent separation. In the overstability anomaly precursor sequence, the time period w(k) carrying the cascaded instability label is set to 1.5 to ensure that the magnitude of the separation risk index matches the actual risk intensity of synchronous low-pulsation in multiple zones. The w(k) for low-confidence labeled areas is reduced to 0.7 to prevent sensor estimation errors from amplifying low-quality signals into artificially high risk levels. A synchronous rise in DRI(i,t) in adjacent zones indicates that flow separation is spreading spanwise. An isolated high DRI(i,t) in a single zone while the surrounding area remains normal usually corresponds to early local separation rather than overall flow field instability. These two types of patterns can be directly distinguished by the spanwise distribution of the separation risk index.

[0024] In some embodiments, the step of identifying latent separation regions by screening long-term pressure stability characteristics based on the separation risk index includes: calculating a regional stability index by performing a time-domain stability index on the separation risk index; performing a long-term continuity test on the regional stability index to identify ultra-long stable segments; performing multi-region synchronous low-pulsation collaborative identification based on the ultra-long stable segments to generate a synchronous stable anomaly map; and using the synchronous stable anomaly map to identify separation diffusion sources and determine latent separation regions.

[0025] The regional stability index is obtained by calculating the time-domain stationarity index of the separation risk index. The regional stability index is calculated by dividing the mean of DRI(i,t) of each region within the sliding window by the standard deviation, as follows: PSI(i,t)=μ_DRI(i,t) / σ_DRI(i,t), where PSI(i,t) is the regional stability index of region i at time t; μ_DRI(i,t) is the mean of the separation risk index of region i within the sliding calculation window at time t; σ_DRI(i,t) is the standard deviation of the separation risk index within the same window. When σ_DRI approaches zero, a minimum value is set to prevent division by zero. The minimum value is determined in the initialization stage based on the lower limit of the historical normal navigation period separation risk index fluctuation of each region. A larger PSI(i,t) indicates a narrower fluctuation in DRI(i,t) over the time domain and a more stable risk signal. When DRI(i,t) is consistently high in a certain region and the fluctuation amplitude narrows synchronously within the window, μ_DRI(i,t) rises while σ_DRI(i,t) falls, resulting in a significant amplification of PSI(i,t). This corresponds to the flow separation in that region having transitioned from an occasional disturbance to a persistent locked-in state. In the mid-sail suction region, after a sustained deflection of the wind direction, DRI(i,t) no longer oscillates slightly with the incoming turbulence but maintains a nearly constant output at a high level, and PSI(i,t) rises rapidly. Meanwhile, DRI(i,t) in the regions at both ends of the sail continues to oscillate normally. The difference in spanwise PSI(i,t) precisely pinpoints the spatially locked-in location of the flow separation. After a brief isolated spike in the separation risk index time series, σ_DRI(i,t) is amplified while μ_DRI(i,t) is only slightly affected. PSI(i,t) is actually lower after such brief anomalies, demonstrating a natural filtering effect on isolated disturbances and preventing misjudgment of a single hyperstable event as a long-term stable trend. The calculation window length is determined during the initialization phase based on the evolution period of the sail flow separation. Too short a window results in excessive sensitivity to brief disturbances, while too long a window leads to a delayed response to newly emerging sustained stable trends. The PSI(i,t) calculation results in the low-confidence marked region are directly reduced by a factor of 0.7 to prevent low-quality signals from interfering with subsequent determinations of ultra-long stable sections.

[0026] Long-term continuity tests are performed on regional stability indicators to identify ultra-long stable segments. Segments where regional stability indicators continuously exceed a threshold and the number of consecutive frames reaches the minimum duration limit are identified as ultra-long stable segments. The start and end frame coordinates are accurate to the sampling period granularity. The mean of regional stability indicators within the segment is synchronously archived for subsequent collaborative strength calculations. The threshold is determined by referring to the 90th quantile of regional stability indicators in historical normal navigation data. Isolated segments where regional stability indicators exceed the threshold but the number of consecutive frames does not reach the minimum do not constitute ultra-long stable segments; only a short-term stability warning is recorded. Short-term warnings and ultra-long stable segments are stored separately in the time-series records. The former is merely an observation clue, while the latter is the identification result. They have different roles in the multi-area synchronous collaborative identification stage and should not be treated interchangeably. Segments where the regional stability index spikes briefly and then falls back as the incoming flow direction recovers often occur in segments where the incoming flow undergoes a phased deflection. The duration of these segments usually falls below the lower limit. Short-term stability warnings are only used as a reference, not a trigger condition, in collaborative identification; therefore, phased disturbances in the navigation environment will not accumulate into ultra-long stable segments. The minimum duration lower limit is taken from the shortest duration of the preceding state of actual flow separation in historical data. Stable segments shorter than this duration are not considered deeply stable, and including them in the identification process would only introduce noise. Intermittent crossings within ultra-long stable segments follow the same tolerance rules as ultra-stable segments with long pressure differentials: the number of crossing frames remains continuous within the tolerance; otherwise, they are truncated in place into two independent ultra-long stable segments. When the interval between two adjacent ultra-long stable segments in the same region does not exceed the set merging window (which can be half of the minimum duration lower limit), they are merged into a single segment with a longer duration. The merged ultra-long stable segment is more likely to meet the time overlap determination condition during the multi-region synchronous collaborative identification phase.

[0027] Based on ultra-long stable sections, multi-zone synchronous low-pulsation collaborative identification is implemented to generate synchronous stable anomaly maps. Ultra-long stable sections in each zone are aligned and compared on the time axis. Multiple ultra-long stable sections with a time overlap exceeding a set ratio and adjacent spanwise trigger synchronous collaborative recording. The set of synchronously triggered regions and their corresponding time periods constitute a collaborative unit of the synchronous stable anomaly map. The strength of the collaborative unit is weighted and quantified by the mean of the stable indicators archived in the ultra-long stable sections of the participating regions. When high angle-of-attack conditions are maintained for a long time and the wind speed is relatively stable, ultra-long stable sections in multiple spanwise zones will appear synchronously due to the weakening of overall inflow attachment. The synchronous stable anomaly map distinguishes this overall effect from local separation diffusion based on the difference in regional stable indicators of each zone. The difference can be taken as the ratio of the range to the mean of the regional stable indicators of each zone within the collaborative unit: a low ratio indicates that the entire sail surface is flattened by the same inflow factor, and the collaborative unit is marked with an overall inflow label; a high ratio indicates that some areas are significantly deeper than others, corresponding to local separation dominance. The two types of labels follow different processing paths in the hidden separation zone identification stage. Long, isolated, stable segments do not constitute cooperative units and are stored separately in the synchronous stable anomaly map as single-region identifiers. Single-region and multi-region cooperative entries are listed separately in the synchronous stable anomaly map, allowing for rapid risk level differentiation in subsequent separation and diffusion source identification. Synchronization overlap is determined by scanning the alignment relationship of each region using a sliding time window: the overlap threshold is relatively wide when two regions are synchronized, such as 60% triggering a strong cooperative diffusion event; it is tightened to around 75% when three or more regions are synchronized. Occasional temporal proximity is not considered a strong cooperative diffusion event. High cooperative strength indicates low-pulsation depth synchronization across multiple regions; the implicit separation regions corresponding to these units undergo key verification using differential pressure gradients during the partition risk map stage.

[0028] Synchronous stationary anomaly maps are used to identify separation and diffusion sources and determine hidden separation zones. The start times of the longest stationary segments within each coordinating unit of the synchronous stationary anomaly map are ordered longitudinally, with the earliest starting region identified as the separation and diffusion source for that unit. The source, along with one or two directly adjacent downstream sub-regions along the diffusion direction, is collectively defined as the hidden separation zone. The longitudinal diffusion direction is determined by the gradient direction of the start time, i.e., the side where the start time increases longitudinally. Under high angle-of-attack conditions, the attachment conditions in the middle leeward region are inherently weakest. In the same flight, these regions often appear as coordinating unit sources in multiple consecutive statistical windows, and the stationary anomalies in the surrounding areas progressively advance longitudinally. Regions that repeatedly act as diffusion initiation points are marked as persistent separation core areas in the hidden separation zone. Repeated source markings are added when three or more consecutive statistical windows trigger the same event. These repeated source markings serve as the basis for determining the highest risk level during the sub-region risk map generation stage. In the synchronous stationary anomaly map, the cooperating units carrying overall incoming flow labels have their source identification results appended with incoming flow disturbance labels in the latent separation region. These are managed separately from the latent separation region entries dominated by actual separation. The overall stationarity caused by the weakening of the incoming flow should not be regarded as the source of separation and diffusion. Individual region labels in the synchronous stationary anomaly map, verified through historical comparison, indicate that these regions have historically appeared as sources of cooperating units. Current single-region triggering is considered an early signal of a new round of separation accumulation, recorded separately with a latency period label. Regions with latency period labels occupy the second highest risk level in the zonal risk map, suggesting a historical precedent of evolving into multi-regional cooperating patterns.

[0029] A zonal risk map is established by tracking changes in pressure gradient between adjacent regions along the span of a hidden separation zone. The regional pressure gradient sequence of each zone within the hidden separation zone and its adjacent regions along the span is subtracted across zones over time periods and divided by the spanwise distance to obtain the spanwise pressure gradient intensity. When the gradient intensity is consistently high and its direction points inward into the hidden separation zone, the risk level of the corresponding segment in the zonal risk map is upgraded by one level. The monotonically increasing spanwise pressure gradient over time indicates a tendency for the hidden separation zone to expand outward. The window-by-window increase in gradient intensity at the boundary is a direct indicator of the expansion rate; a higher increase corresponds to faster expansion. During adjustments with increasing angle of attack, the pressure gradient at the edge of the hidden separation zone often rises sharply within seconds. When the abrupt change exceeds the historical 90th quantile, a rapid expansion marker is added to the corresponding cell in the zonal risk map. This rapid expansion marker is a typical trigger feature for identifying rapid expansion of the separation zone. When the gradient intensity drops sharply and the differential pressure at the corresponding location of the latent separation zone returns to normal oscillation, the risk label of the corresponding time period on the zoning risk map is adjusted downwards accordingly. If the gradient decreases but the separation risk index remains high, the separation is more likely to shift to a deeper location rather than a true mitigation, and the risk level remains unchanged. The zoning risk map is constructed with the zoning zone number as the row and the time window as the column, and it is updated as the statistical window scrolls. The cell stores the comprehensive risk level of the corresponding area during that time period. The level is synthesized by normalizing the separation risk index and the differential pressure gradient intensity separately and then according to the set weights. The persistent core area of ​​the latent separation zone corresponds to the highest risk level, the area marked with the incubation period corresponds to the second highest level, and the area marked with incoming flow disturbance corresponds to the baseline level. The three levels are recorded hierarchically in the zoning risk map. During the differential pressure coordinated adjustment phase, high-priority intervention areas can be directly located based on the zoning risk map.

[0030] Step S103: Based on the risk map of the partition, perform weighted iterative fitting of sudden change in the angle of attack of the incoming flow to obtain the partition negative pressure setpoint. Implement overload-triggered high-confidence calibration on the partition negative pressure setpoint to construct a self-correcting mapping table. Verify the historical overload distribution of the self-correcting mapping table to form a dynamic calibration mapping.

[0031] Specifically, the given value of negative pressure in each zone is obtained by performing a weighted iterative fitting of the sudden change in the incoming flow angle based on the risk map of the zone. Historical angle-of-attack change events and negative pressure response samples for each zone are read from the operation records zone by zone. Sample points within the current angle-of-attack operating range are used for fitting and calculated using the following formula: P_set(i)=Σ_j[w_t(j)×w_r(i,j)×P_obs(i,j)] / Σ_j[w_t(j)×w_r(i,j)], where Σ_j represents the summation of all historical sudden change events j within the current angle-of-attack range, P_set(i) is the given negative pressure value for zone i; P_obs(i,j) is the measured negative pressure value of zone i during the j-th historical sudden change event within the current angle-of-attack range; w_t(j) is the time decay weight of the j-th event, assigned according to exponential decay, with recent events having a higher weight than distant events; w_r(i,j) is the risk level weight of zone i in the risk map during the j-th event, with samples from high-risk periods having a higher weight than those from low-risk periods. After adjusting the sample weights according to the fitted residuals in each round, the calculation is recalculated. The iteration terminates when the change between two adjacent rounds of P_set(i) is less than the set convergence threshold. The convergence value is the negative pressure given value for that zone. When there are insufficient samples in the current angle of attack zone, samples from adjacent angle of attack zones are used to supplement the negative pressure given value. The w_r of the region marked with rapid expansion in the risk map is increased by 1.3 times to reflect the higher reference value of the zone negative pressure given value during the rapid expansion period of the separation zone. The w_r of the region marked with incoming flow disturbance is reduced to 0.7 to prevent the low negative pressure samples during the period of overall weakening incoming flow from systematically suppressing the zone negative pressure given value. When ships navigate around islands or enter narrow channels, the angle of attack of the sails often changes abruptly by more than 5 degrees within seconds. The w_t(j) of such sudden events is high due to the short time interval between their occurrences. The risk level of the corresponding regional risk map is usually at a high level simultaneously, which makes the w_r(i,j) superimposed and higher. The double weighting amplifies the occurrence of such high-pressure sudden events, making them dominant in the fitting of P_set(i). Therefore, the negative pressure setpoint of the region is given a more conservative negative pressure preset under the current similar inflow conditions to avoid the lag in the negative pressure response of each region when the sudden change occurs.

[0032] In some embodiments, the step of constructing a self-correcting mapping table by implementing overload-triggered high-reliability calibration of the partition negative pressure setpoint includes: real-time monitoring of the operating status of the partition negative pressure setpoint to obtain fan operating status quantities; screening for excessive current triggering periods based on the fan operating status quantities to establish an overload excitation point set; extracting synchronous operating condition parameters based on the overload excitation point set to obtain an overload calibration benchmark library; and performing iterative weighted fusion of the overload calibration benchmark library to form a self-correcting mapping table.

[0033] Real-time monitoring of the fan operating status is performed on the zone negative pressure setpoint to obtain fan operating status data. Each update of the zone negative pressure setpoint triggers the status acquisition of the corresponding zone's fans. Rotational speed, effective values ​​of three-phase current, and bearing vibration acceleration are synchronously written into the timing buffer of each zone at a sampling rate higher than the setpoint update frequency. The sampling rate can be more than five times the update frequency. These three types of parameters are time-aligned and arranged side-by-side to constitute the fan operating status data. The real-time changes in rotational speed and three-phase current are the core basis for subsequent excessive current screening. After the zone negative pressure setpoint is sent to the fan drivers of each zone, the actual execution speed fed back by the drivers is written into the fan operating status data, serving as the basis for calculating the setpoint following error. When the following error exceeds a set percentage (e.g., 5%) for multiple consecutive update cycles, the corresponding entry is marked with a following anomaly. When a district experiences a sudden increase in negative pressure setpoint, requiring a significant speed increase of the fan within a short period, the driver current will exhibit a brief surge followed by a plateau in the initial stage of speed increase. This current plateau is a typical precursor to overload excitation identification. The high sampling rate of the fan operating status variables is designed to capture this plateau, as it is easily missed by the entire plateau segment at a low sampling rate. Periods in the fan operating status variables where the vibration acceleration exceeds a set threshold are recorded as vibration warnings. The simultaneous occurrence of vibration warnings and current overload generally indicates that the fan's mechanical state is approaching the overload boundary rather than experiencing normal load fluctuations. The temporal overlap of these two types of markings serves as an auxiliary criterion for identifying the true source of overload during the overload time sequence marking stage. When a sensor experiences a brief disconnection, resulting in missing frames in the fan operating status variables, linear interpolation of the preceding and following valid frames is used to replace them. The interpolated frames are marked with interpolation labels and do not participate in the triggering determination of overload current. Interpolated values ​​are not actual measurements; triggering based on them constitutes a false signal. Samples with interpolated frames in the plateau segment are marked with an interpolation contamination label.

[0034] For example, the step of establishing an overload excitation point set for screening the excessive current triggering period of the fan operating status quantity includes: performing multi-phase current decomposition on the fan operating status quantity to obtain phase current records; marking the phase current records with the current plateau segment before overload to obtain an overload time sequence; performing synchronous operating condition parameter capture based on the overload time sequence to establish an overload excitation sampling library; and performing excitation point aggregation analysis based on the overload excitation sampling library to establish an overload excitation point set.

[0035] Multi-phase current decomposition is performed on the fan operating status variables to obtain phase-by-phase current records. The instantaneous values ​​of the three-phase current carried by the fan operating status variables are extracted independently for phases A, B, and C. The effective values ​​of each phase current are calculated using a fixed-frame sliding window and then arranged side-by-side according to phase number. The amplitude difference of the three-phase effective values ​​and the phase imbalance are synchronously written into the corresponding fields to form the phase-by-phase current record. The phase imbalance can be taken as the ratio of the maximum deviation of the three-phase effective values ​​to the three-phase average. During the stage of sudden negative pressure rise in the suction sail fan, due to the phase control accuracy of the driver, a phenomenon of single-phase current exceeding the limit often occurs. At this time, the synthesized effective value is still within the normal range, but a single phase has already reached the overload boundary. Phase-by-phase decomposition can capture the single-phase overload signal several frames earlier than the synthesized average value in the fan operating status variables, thus shifting the trigger time of the overload time sequence forward and leaving a larger response margin for overload protection. The instantaneous current values ​​of each phase corresponding to the vibration warning annotation frames in the fan operation status data are synchronously inherited from the vibration warning annotations when written into the phase current record. This allows the overload time sequence annotation stage to distinguish between current fluctuations caused by mechanical vibration and active current rises caused by negative voltage regulation. Current fluctuations caused by mechanical vibration often exhibit high-frequency periodic jitter, while active rises caused by negative voltage regulation show a monotonous climb followed by a stable plateau. The difference between these two waveforms enables the phase current record to effectively identify the source of excess current in the subsequent plateau segment annotation stage. Frames in the phase current record where the phase imbalance exceeds a set threshold are written with an imbalance annotation. These imbalance annotation frames trigger single-phase excess priority judgment during the overload time sequence identification stage. Under imbalance conditions, the three-phase average will average out the actual excess of a single phase, and single-phase priority judgment blocks this missed detection path. Prolonged high-power pumping increases the driver's heat dissipation pressure, and the phase control accuracy decreases with rising temperature. The density of imbalance annotation frames in the phase current record remains high during these periods, which is an early signal for identifying driver thermal degradation.

[0036] The overload time sequence is obtained by annotating the phase current records with pre-overload current plateau segments. The effective values ​​of each phase in the phase current records are scanned frame by frame. The segment where the current remains between 80% and 95% of the rated value for several consecutive frames before overload triggering, and where the amplitude change rate is lower than a set slope threshold, is marked as the pre-overload current plateau segment. The frame where the effective current value exceeds a multiple threshold (e.g., 1.2 times the rated value) after the plateau segment ends is defined as the overload moment. All overload moments are arranged chronologically to form the overload time sequence. When the fan current surges directly from a low value without a plateau transition during a sudden increase in no load, this surge corresponds to the driver's own protection action rather than the actual operating limit caused by negative pressure regulation. Overload moments supported by a plateau segment reflect the fan's gradual approach to its limit under high load, and their calibration value is far higher than the surge triggering without a plateau segment. The two types of overload moments are distinguished in the overload time sequence by the presence or absence of a plateau segment annotation. Both widening and narrowing the slope threshold have their costs: if it's too wide, the slow climb phase might be misjudged as a plateau, and the surge trigger might be incorrectly labeled with a plateau segment; if it's too narrow, some real plateau segments might be missed. The value can be determined by referring to the measured slope distribution of the effective current values ​​within the 80% to 95% rated range in the historical data of the phase current records. For conservatism, the threshold value should not be lower than its 95th percentile. The overload moment corresponding to the unbalanced label frame in the phase current record is appended with a single-phase overload indicator to the overload moment sequence. The single-phase overload indicator moment is recorded synchronously with the trigger phase field during the overload excitation sampling library construction phase, thus completely preserving the real overload condition of a single phase. When the interval between two adjacent overload moments in the overload moment sequence is less than the set merging window, they are merged into a single continuous overload event. The merging window length is calculated as the number of sampling frames based on the driver thermal protection recovery time, with the first overload moment as the start point and the last as the end point. When constructing the overload excitation sampling library, the operating condition parameters are extracted as a whole from the merged event, and the same overload process is no longer divided into multiple independent events for repeated labeling.

[0037] An overload excitation sampling library is established by synchronously capturing operating condition parameters based on the overload time sequence. The time marked with a platform segment in the overload time sequence is used as the backtracking starting point. One operating condition acquisition cycle is traced back to extract the incoming flow angle of attack, ship speed, measured negative pressure in each zone, and fan speed. These four types of operating condition parameters, along with the trigger phase, peak current, and platform segment duration of the current overload time, constitute the operating condition sample row for that overload event. All sample rows are then aggregated to form the overload excitation sampling library. When the incoming flow intensity repeatedly and rapidly increases within a short period, overload times within the same operating condition interval appear densely in the overload time sequence, resulting in a sharp increase in the sample size of the corresponding interval in the overload excitation sampling library. The peak currents of these concentrated trigger samples are close to each other, avoiding unstable weighting processing during the overload excitation point set aggregation and analysis stage, and their calibration confidence is higher than that of sample groups with occasional and scattered triggers. Multiple overload times in the overload time sequence that are merged into a continuous event share a set of operating condition parameters in the overload excitation sampling library, using the backtracking value before the start time of the merged event as the standard. During continuous overload periods, no further capture is performed. In the overload excitation sampling library, when the peak current dispersion of multiple sample rows within the same operating condition interval exceeds a set threshold, an unstable operating condition label is added. The dispersion can be taken as the coefficient of variation of the peak current of the sample in that interval. Samples with unstable operating conditions are subject to weight reduction during the overload excitation point set aggregation and analysis phase, since the current peak during the operating condition fluctuation period does not represent the steady-state excitation intensity. The unstable operating condition label is most common in sections with large speed changes: when the ship speed changes suddenly, the flow pressure fluctuates greatly during short periods, and sample rows from multiple operating condition intervals will be labeled as unstable at the same time. This batch weight reduction does not affect the calibration quality of the overload excitation point set in the normal navigation operating condition interval. The overload excitation sampling library distinguishes between normal operating conditions and dynamic transition operating conditions by the ship speed change rate field. Periods with a change rate exceeding a set threshold are additionally labeled as dynamic operating conditions, and are weighted at 0.6 times during aggregation.

[0038] An overload excitation point set is established based on the overload excitation sampling library and excitation point aggregation analysis. Sample rows belonging to the same operating condition range are arranged in descending order of peak current. Samples exceeding the mean plus one standard deviation of all samples in that operating condition range are marked as high-intensity excitation points, while the rest are marked as ordinary excitation points. These two types of excitation points are stored separately according to their intensity level, together forming the overload excitation point set. When a certain high-load operating condition is repeatedly experienced during operation, the peak current samples of the same operating condition range in the overload excitation sampling library continuously accumulate, and the proportion of high-intensity excitation points in the corresponding range of the overload excitation point set increases accordingly. The high-intensity proportion is the ratio of high-intensity excitation points to the total number of excitation points in that range. This characteristic triggers faster step-size convergence during the self-correcting mapping table fusion stage, appropriately tightening the upper bound of the negative voltage for that operating condition range to match the actual bearing limit of high-frequency overload scenarios. Unstable samples in the overload excitation sampling library are weighted at 0.6 times in the overload excitation point set; the weighting process in the aggregation analysis stage is thus implemented as a specific reduction factor. When the sample size for a certain operating condition interval is less than three, the weighted average of high-intensity excitation points in adjacent operating condition intervals is used to estimate and add low-sample annotations. The low-sample annotation excitation points, along with the estimated operating condition parameters, are directly written into the overload calibration benchmark library and included in the fusion process with a step size of 0.075. When samples are scarce, it is preferable to estimate the operating condition limit more conservatively. Operating condition intervals with a high-intensity proportion of overload excitation points that continuously increases across multiple voyages often indicate long-term degradation of the aerodynamic performance of the sail in that region. When the high-intensity proportion of the same operating condition interval exceeds the global median for three consecutive voyages, the overload excitation point set adds performance degradation precursor annotations to that interval. The performance degradation precursor annotation interval triggers early boundary narrowing processing during the dynamic calibration mapping verification stage. Conservative adjustments are initiated before the short-term mutation annotation conditions are met, which can provide a quantitative basis for sail maintenance priority ranking.

[0039] An overload calibration benchmark library is obtained by extracting synchronous operating condition parameters based on the overload excitation point set. The start time of each excitation event in the overload excitation point set is used sequentially as a backtracking anchor point. The incoming flow angle of attack, ship speed, measured negative pressure in each zone, and fan speed stored in the overload excitation sampling library at this anchor point are reused. These four types of operating condition parameters, together with the trigger phase, peak current, and platform segment duration of the current overload event, constitute the calibration sample row for that event. All sample rows are collected to form the overload calibration benchmark library. The length of the backtracking time window follows the delay in the transmission of changes in the incoming flow angle of attack to the fan current response: a shorter window will result in the captured operating condition parameters not yet stabilizing in the current overload state; a longer window will mix in historical operating condition parameters from before the overload, thus contaminating the calibration benchmark. When overload excitation in a certain area is repeatedly triggered under the same angle of attack condition, the overload excitation point set accumulates a large number of samples in a short period within that operating range, causing a sharp increase in the sample size of the corresponding range in the overload calibration benchmark library. These concentrated triggers are distinguished by high-frequency annotations and ordinary intermittent operating conditions. The judgment threshold can be set at more than twice the median of the entire range. High-frequency annotated ranges trigger faster step-size convergence during self-correcting mapping table fusion, causing the upper limit of negative pressure in that operating range to be quickly lowered to a safe range. Events carrying follow-up abnormal annotations during overload periods are stored hierarchically in the overload calibration benchmark library along with normal overload excitation events. The fusion weight of samples in the normal overload layer is higher than that in the follow-up abnormal layer because the former directly reflects the true correspondence between the given negative pressure value of the zone and the actual load limit of the fan. The ratio of the sample size of the normal overload layer to the following abnormal layer varies systematically across different operating conditions. In the range where the angle of attack changes frequently, the sample size of the normal overload layer is usually much larger than that of the following abnormal layer. In the range of long-term high-load stable conditions, the sample sizes of the two layers are relatively balanced. The hierarchical storage of the overload calibration benchmark library makes the contributions of overload events of the two mechanisms to the correction of the self-correction mapping table independent and traceable.

[0040] The overload calibration benchmark library is iteratively weighted and fused to form a self-correcting mapping table. After sorting the calibration samples of each operating condition interval in the overload calibration benchmark library according to the time-reverse decay weight, they are iteratively fused with the corresponding operating condition entries of the current version of the self-correcting mapping table according to the following formula: M_new(i,α,v)=(1-λ(j))×M_old(i,α,v)+λ(j)×P_lim(i,j), where M_new(i,α,v) is the updated entry value of the self-correcting mapping table in the i-th region under the operating conditions of the incoming flow angle of attack interval α and the ship speed interval v; M_old(i,α,v) is the entry value before the update; λ(j) is the fusion step size of the j-th time, which is 0.15 for the normal overload layer, 0.10 for the following abnormal layer, and 0.075 for the interpolated contamination and low sample labeled samples; P_lim(i,j) is the upper limit observation value of the negative pressure corresponding to the i-th region in the j-th sample of the overload calibration benchmark library, which is obtained by back-calculating the peak current of the sample row according to the fan current-negative pressure characteristic curve. When historical overload excitation events occur frequently in a certain operating condition range, the sample size in this range is large and the time weight of recent samples is relatively high. After multiple iterations, M_new(i,α,v) quickly converges to the upper limit of the safe negative pressure under this operating condition. When the subsequent partition negative pressure setpoint exceeds this upper limit, the self-correcting mapping table automatically triggers a limiting correction to prevent the setpoint from continuously exceeding the actual load capacity of the fan. In the overload calibration benchmark library, the λ(j) corresponding to the two types of low-confidence samples mentioned above is 0.075. The low-confidence samples are gradually diluted rather than completely eliminated under the joint drive of multiple normal samples, and the value of supplementary samples is still retained in the extreme boundary range of the operating condition. After the self-correcting mapping table converges and stabilizes in each operating condition range, the corresponding upper limit of negative pressure is directly used as the limiting constraint of the negative pressure setpoint of the next batch of partitions, so that the negative pressure adjustment range of the suction sail under known overload risk operating conditions is substantially narrowed, and the systemic overload risk continues to decrease with version iterations.

[0041] A dynamic calibration map is formed by verifying the historical overload distribution of the self-correcting mapping table. The negative pressure upper bound values ​​stored in each operating condition interval of the self-correcting mapping table are verified zone by zone against the historical overload excitation frequency distribution of the same operating condition interval. Intervals with frequencies higher than the global median are marked as high-frequency overload intervals, and those lower than the global median are marked as low-frequency overload intervals. The verified and adjusted negative pressure upper bounds for each zone, along with the interval labels, are stored using an angle-of-attack-ship speed operating condition index, constituting the dynamic calibration map. The negative pressure upper bound of the high-frequency overload interval is lowered in the dynamic calibration map by a set step size (e.g., 2% per batch). This operating condition interval has been corrected by overload events for a long time, indicating that the original upper bound of the self-correcting mapping table was overly optimistic. The negative pressure upper bound of the low-frequency overload interval is slightly raised after several consecutive batches without overload records, releasing the negative pressure adjustment space that has been underutilized in this operating condition interval for a long time. When the overload frequency of a certain operating condition range surges in recent consecutive batches while historically the long-term frequency remains low, the dynamic calibration map adds short-term abrupt change markers to this range. This combination suggests that the overload excitation originates from recent changes in sail or actuator status, rather than long-term operating condition characteristics. The dynamic calibration map uses a shorter time decay window for the short-term abrupt change and performance degradation precursor marker ranges, allowing the recent correction effect to be quickly reflected, and historical low-frequency data cannot suppress the recent true deterioration trend. The boundary adjustment of the short-term abrupt change marker range is accelerated due to the narrowing decay window. In the early stages of sail aerodynamic performance degradation, the upper limit of the negative pressure in this range can be conservatively narrowed quickly within several batches, completing boundary updates 3 to 5 batches earlier than the conventional iteration of the self-correcting mapping table. The partition negative pressure setpoint is effectively constrained in the early stages of performance degradation.

[0042] Step S104: Implement the separation and shielding full-area differential pressure coordination regulation output differential pressure regulation scheme based on dynamic calibration mapping, and perform graded under-pumping boundary approximation based on differential pressure regulation scheme to obtain the attached flow distribution map.

[0043] Specifically, a differential pressure regulation scheme based on dynamic calibration mapping is implemented, employing a separation-shielding, full-area differential pressure coordinated regulation output differential pressure control method. The upper limit of negative pressure in each area under the current operating condition index is read from the dynamic calibration mapping to construct a regulation constraint matrix. Coordinated regulation commands within the constraint matrix independently calculate differential pressure regulation amounts for each area. High-risk areas on the zoning risk map are marked with separation-shielding identifiers in the regulation constraint matrix. The negative pressure command for the shielded area is locked at its current value during the current regulation cycle, without any upward adjustment. The differential pressure regulation scheme outputs the full-area regulation set and the shielded area identifiers. Assuming separation-shielding identifiers are already displayed in the two mid-section areas under high angle-of-attack conditions: Applying a large negative pressure step to the shielded area at this time can easily induce adsorption boundary layer oscillation instability, instead accelerating flow separation and spreading to adjacent areas. The separation-shielding mechanism locks the two mid-section areas and guides coordinated compensation in the low-risk end areas. The upward adjustment step size of the unshielded area is allocated within the dynamic calibration mapping constraint framework according to the proportion of the upper limit margin of each area, keeping the overall lift loss within a safe threshold (e.g., not exceeding 3% of the rated total lift). The low-frequency overload region in the dynamic calibration mapping receives a relatively larger adjustment step size in the coordinated adjustment, while the step size in the high-frequency overload region is conservatively compressed. The difference in adjustment amount between the two types of regions in the differential pressure adjustment scheme directly reflects the constraint of historical overload distribution verification results on current adjustment behavior. Even if the regions marked with latent periods in the zoning risk map do not reach the high-risk level, the adjustment step size in the differential pressure adjustment scheme is also subject to additional conservative restrictions. Adjusting close to the upper limit would easily trigger the potential separation risk of these regions, so the corresponding adjustment amount is kept within 80% of the upper limit of the dynamic calibration mapping constraint. The adjustment step size in the region marked with short-term mutations in the dynamic calibration mapping is further tightened to 60% of the normal step size. The historical negative pressure upper limit of this region is rapidly shifting downward, so the adjustment is conservatively implemented first to prevent the differential pressure adjustment scheme from exceeding the actual safety boundary before the mapping convergence is completed.

[0044] In some embodiments, the step of obtaining the attached flow distribution map by performing graded under-pumping boundary approximation based on the differential pressure regulation scheme includes: quantifying the lift redundancy of each region based on the differential pressure regulation scheme to obtain a lift redundancy graded table; performing under-pumping timing planning based on the lift redundancy graded table to determine an under-pumping timing table; performing critical attachment limit approximation on the under-pumping timing table to generate a partitioned execution control quantity; and using the partitioned execution control quantity to confirm the lower limit safety boundary of attachment to obtain the attached flow distribution map.

[0045] Based on the differential pressure regulation scheme, the lift redundancy of each zone is quantified to obtain a lift redundancy classification table. The actual executed negative pressure value of each zone under the differential pressure regulation scheme is normalized after subtracting the historical critical attached negative pressure of that zone, and the lift redundancy of each zone is obtained by the following formula: L_rem(i)=(P_cur(i)-P_min(i)) / P_cur(i), where L_rem(i) is the lift redundancy of the i-th zone, taking a value from 0 to 1; P_min(i) is the critical value that triggers the separation alarm during the historical negative pressure decline in that zone; P_cur(i) is the current executed negative pressure value of the i-th zone under the differential pressure regulation scheme; the larger L_rem(i) is, the more adjustment space is available in that zone based on the current negative pressure, and L_rem(i) approaching 0 indicates that the zone has approached the historical safety limit. Each zone's L_rem(i) is divided into three levels: high (L_rem≥0.4), medium (0.2≤L_rem<0.4), and low (L_rem<0.2), which, together with the zone number, constitute a lift redundancy classification table. In the differential pressure adjustment scheme, the P_cur(i) of the separated shielded area is replaced by the most recent effective negative pressure value before shielding. The corresponding L_rem(i) is marked with additional shielding estimation and is placed in a separate layer in the lift redundancy classification table. During the under-pumping timing planning stage, the shielding estimation marked area adopts the most conservative step size to prevent estimation deviation from causing the area to reach the critical point prematurely. When the ship is fully loaded, the waterline is higher, which reduces the effective windward height of the sail. In most pumping zones, P_min(i) systematically increases under full load conditions. The number of low redundancy zones in the lift redundancy classification table is significantly higher in the full load batch than in the ballast state. Under-pumping timing planning needs to tighten the upper limit of the step size as a whole to prevent the lift from dropping sharply beyond the safety boundary when multiple zones approach the critical point simultaneously. When the low-redundancy zone in the lift redundancy classification table highly overlaps with the high-risk zone in the zoning risk map, it indicates that the overall negative pressure adjustment margin of the sail is insufficient. The lower limit safety boundary of attachment generated by the subsequent attachment flow distribution map is more conservative in this type of batch than in normal batches.

[0046] Based on the lift redundancy classification table, under-pumping timing planning is conducted to determine the under-pumping timing table. The redundancy level of each zone in the lift redundancy classification table determines the allowable step size and execution cycle of the under-pumping operation: high redundancy zones allow larger step sizes to quickly approach the critical point, while low redundancy zones only allow fine step sizes for step-by-step detection. The operation cycles of the two types of zones are staggered in the under-pumping timing table, with high redundancy zones being tested first and low redundancy zones following for verification. This prevents the total lift from fluctuating significantly in the short term due to simultaneous under-pumping in multiple zones. The under-pumping step size is subject to two constraints: the upper limit is the detection resolution; if the step size exceeds the resolvable interval near the critical point, the critical point will be skipped in one step. The lower limit is the timing efficiency; if the step size is too fragmented, the execution time of the under-pumping timing table will be prolonged, encroaching on the normal negative pressure regulation cycle. The value is taken as a reference to the minimum detection step size triggered by the historical critical point of each redundancy level. When there is a significant difference in redundancy levels between adjacent areas, the lift redundancy classification table triggers a step size protection strategy for adjacent areas, ensuring that the step size difference between two adjacent areas in the same execution round does not exceed the set upper limit. Large step size movements in high redundancy areas can affect adjacent low redundancy areas through spanwise pressure difference coupling waves, and the upper limit prevents the detection disturbances that should be limited to a single area from being contained within that area. In the under-pumping timing table, the execution cycle of the shielded estimation marked area in the lift redundancy classification table is scheduled after all unshielded areas have completed at least one round of detection. The estimation deviation is corrected with measured data before the under-pumping test in the shielded area is started. In the first batch after maintenance, the P_min(i) of some areas is lowered due to component updates. The L_rem(i) of these areas in the lift redundancy classification table is temporarily raised to the high redundancy level. In this batch, the under-pumping timing table prioritizes scheduling them to complete the critical point recalibration, so that the attached flow distribution map is quickly updated to the new safety boundary benchmark after maintenance.

[0047] The under-pumping timing table is approximated to the critical attachment limit to generate partition execution control quantities. The negative pressure command for each zone is progressively reduced according to the partition execution rhythm specified in the under-pumping timing table. After each reduction, a specified number of stable frames are waited before evaluating the current attachment state. If the pressure difference stability measurement in each zone maintains normal oscillation within the stable frames, the next step continues. If the pressure difference stability measurement exceeds the set alarm threshold, the execution of that step is immediately stopped, and the negative pressure command value of the previous step is recorded as the critical candidate value for this round of detection in that zone. The partition execution control quantity is output with the set of critical candidate values ​​for all zones as its core. The number of stable frames directly affects the accuracy of critical point identification: too few frames result in false alarms and premature detection termination due to noise in a single frame; too many frames slow down the detection process and make the overall critical candidate values ​​overly conservative. A compromise is to use twice the duration of typical noise in the pressure difference stability measurement of each zone as the number of frames, balancing response speed and anti-interference capability. When wind speed drops sharply, the incoming flow pressure decreases, increasing the dependence of the attached boundary layer on the suction rate. The critical point in the region originally set for a large step size in the under-suction time series appears earlier than expected. The pressure differential stabilization measurement prematurely exceeds the limit during step execution, triggering termination. Consequently, the critical candidate value recorded by the zonal execution control is lower than the historical average under the same operating conditions. This lower candidate value triggers a one-step backtest during the attached flow distribution map verification phase, thus distinguishing the premature termination caused by dynamic conditions from permanent degradation of attachment capacity. If the measured pressure differential stabilization measurement in the shielded estimation area of ​​the under-suction time series has already triggered an alarm at the initial step size, it indicates that the redundancy estimated before shielding is too high. The zonal execution control marks the critical candidate value of this area with an underestimation correction flag. This underestimation correction flag triggers an additional conservative narrowing of the safety boundary for this area during the attached flow distribution map generation phase.

[0048] The attachment flow distribution map is obtained by confirming the attachment lower limit safety boundary using the partition execution control variable. The critical candidate values ​​for each region of the partition execution control variable are verified region by region against the attachment judgment conditions within the confirmation frame. Verification is based on the number of frames where the pressure difference stability metric is continuously below the alarm threshold, meeting the set stability lower limit. Critical candidate values ​​that pass verification are confirmed as the attachment lower limit safety boundary for that region. The spatial distribution of the lower limit values ​​after confirmation across all regions constitutes the main body of the attachment flow distribution map. Critical candidate values ​​for regions with underestimated correction flags in the partition execution control variable are additionally increased by 3% during the verification stage before participating in confirmation to compensate for the boundary risk introduced by excessively high masking estimation redundancy. Regions that fail verification in the partition execution control variable are backtracked one step and re-executed according to the corresponding step size of the under-sucking timing table. When the number of backtrackings exceeds the set upper limit, the average estimation of neighboring confirmed regions is used instead, and a neighboring estimation label is added to the corresponding position in the attachment flow distribution map. The neighboring estimation label is treated as a low-confidence frame in subsequent attachment shrinkage rate calculations and is only used as a conservative reference during rate setting. The spanwise gradient of the lower limit values ​​of each region in the attached flow distribution map directly reflects the topological structure of the flow separation risk on the sail under the current operating conditions. The location of the gradient abrupt change usually coincides with the location of the high-risk boundary in the partition risk map. A significant deviation between the two indicates that the historical overload distribution verification data of the dynamic calibration mapping has deviated from the current aerodynamic state of the sail. When the deviation exceeds the set threshold, the attached flow distribution map triggers an additional iterative update request for the dynamic calibration mapping, so that the upper limit constraint of the negative pressure in the next batch is realigned with the current measured attached state. The deviation between the historical mapping and the current aerodynamic state will not accumulate and solidify into a systematic conservative error.

[0049] Step S105: Track the contraction rate of the attachment zone using the attached flow distribution map and output a constant speed adjustment set. Identify the lift marginal gain attenuation segment from the speed adjustment set to determine the energy saving target area. Perform active deceleration and overdrive archiving operations on the energy saving target area to generate energy saving control quantity. Verify the energy saving control quantity against global lift constraints and output an electronically controlled negative pressure adjustment command.

[0050] In some embodiments, the step of using the attachment flow distribution map to perform attachment zone contraction rate tracking and constant speed output rotation speed adjustment set includes: performing time-series frame difference calculation on the attachment flow distribution map to obtain the attachment contraction rate; identifying time-by-time stable segments approaching zero based on the attachment contraction rate to obtain an attachment contraction freeze map; establishing a pre-adjusted rotation speed set based on the attachment contraction freeze map by prioritizing constant speed in the frozen continuous extreme value region; and implementing attachment state constraint correction and output rotation speed adjustment set based on the pre-adjusted rotation speed set.

[0051] The attachment shrinkage rate is obtained by temporal frame differential calculation based on the attachment flow distribution map. The lower limit safety boundary value of each region in two adjacent frames of the attachment flow distribution map is subtracted and then divided by the frame interval to obtain the attachment boundary change rate of each region at that moment. The region-by-region rate values ​​at all moments are arranged temporally to form the attachment shrinkage rate. A positive rate value indicates that the attachment boundary is shrinking inwards towards the sail surface, while a negative rate value indicates that the attachment boundary is expanding outwards. Frames with rates adjacent to the estimated marked areas in the attachment flow distribution map are marked with low-confidence rates during calculation. These low-confidence rate frames do not participate in the main sequence of subsequent attachment shrinkage freeze map zero-reaching stability determination, but are retained in the attachment shrinkage rate as an auxiliary reference. Frames with absolute rate values ​​exceeding their historical steady-state fluctuation amplitude by several times (e.g., 5 times) are marked with abnormal rates in the attachment shrinkage rate. These frames mostly correspond to short-term sudden changes in the incoming flow conditions rather than rapid changes in the actual attachment area. During the attachment shrinkage freeze map identification stage, sliding smoothing is performed on several consecutive frames after the abnormal rate marked frames to suppress the impact of sudden change interference on the zero-reaching determination. Short-period shocks from sea states leave regular high-rate pulses in the inter-frame differences of the attachment current distribution map. Abnormal rate markers for the attachment contraction rate appear concentrated during these periods. After smoothing, the rate values ​​return to pre-shock levels, and the identification of the frozen stable segment after the shock remains unaffected. If the abnormal rate marker frame density remains consistently high over multiple consecutive frames rather than as isolated pulses, it indicates a continuous change in the incoming current state. The attachment contraction rate is recorded as a continuous disturbance during this period, and the zero-to-zero threshold for the segment carrying this marker in the attachment contraction freeze map is appropriately tightened, thus reducing the probability of a stable disturbance being misjudged as a true freeze. The frame interval duration remains constant within the normal sampling period. During emergency adjustments, the control system temporarily increases the sampling frequency of the attachment current distribution map. During this period, the rate value must be normalized using the actual frame interval duration, not the nominal period; otherwise, changes in sampling density will directly cause systematic distortion of the attachment contraction rate.

[0052] The attachment-contraction freeze map is obtained by identifying time-by-time stable segments based on the attachment-contraction rate. The rate values ​​of each region are scanned frame by frame. A segment where the absolute value of the rate is below a set zero-reaching threshold for multiple consecutive frames is identified as a stable segment approaching zero. The zero-reaching threshold is the lower quantile (e.g., the 10th quantile) of the steady-state fluctuation amplitude of the historical attachment boundary rate for each region. The start and end frame coordinates of the stable segment approaching zero, together with the corresponding region number, constitute the basic unit of the attachment-contraction freeze map. If the rate value of the frame marked with an abnormal rate in the attachment-contraction rate still exceeds the zero-reaching threshold after smoothing, the current candidate stable segment approaching zero is interrupted and the counting is restarted. This order of smoothing followed by judgment prevents a single sudden change in the incoming flow from mistakenly cutting a real frozen area into two short segments and underestimating its freezing depth. During periods of stable inflow and constant angle of attack, most suction zone attachment boundaries remain essentially stationary under sufficient negative pressure. The attachment contraction rate exhibits a continuous approach to zero over a wide time window during these periods. The number of frozen frames corresponding to these periods in the attachment contraction freeze map is significantly longer. Prioritizing velocity control by using these long, sustained frozen zones as the basis for velocity locking results yields far greater stability than frozen segments that rapidly recover after a brief approach to zero. In the attachment contraction rate, the cells corresponding to persistently disturbed segments are marked with low-confidence freeze labels. Even if the segment's rate value meets the zero-approach condition, an additional number of confirmation frames must be added before freezing can be triggered. This extension increases with the density of persistently disturbed label frames. In this way, the frozen cells in the attachment contraction freeze map will not be artificially increased due to false stabilization caused by brief disturbances.

[0053] Based on the attachment contraction freeze map, a pre-adjusted speed set is established by prioritizing speed control in the extreme freezing regions. Freezing units in each region of the attachment contraction freeze map are arranged in descending order of freezing duration (frames). The region with the longest duration has the highest attachment boundary stability and optimal negative pressure maintenance efficiency; its current fan speed is directly used as a high-reliability speed control anchor point and written into the pre-adjusted speed set. The anchor point speed is taken as the weighted average of the speeds of each frame within the corresponding freezing unit. The weights are determined based on the stability of the lower limit safety boundary of the intra-frame attachment flow distribution map; frames with smaller boundary fluctuations have higher weights. Therefore, the anchor point speed represents the true steady state rather than an instantaneous reading. For regions with shorter freezing durations, the attachment boundary is not yet fully stable. Their corresponding speeds are estimated in the pre-adjusted speed set using a weighted average of the high-reliability anchor points instead of direct values. The instantaneous speed in short-term freeze segments is greatly affected by incoming flow disturbances, and direct values ​​are prone to deviating from the true steady-state target. Short-term freezes often occur when the intensity of incoming flow turbulence weakens in stages. These regions often require higher speeds to maintain attachment after the incoming flow recovers. If the speed is fixed using the short-term freeze speed, the subsequent attachment state may become unstable. Under high angle-of-attack conditions, the leeward edge region typically freezes first and lasts the longest, corresponding to a speed often lower than the overall average. The pre-adjusted speed set established using this region as an anchor point has a downward pulling effect on the overall speed reduction. During the energy-saving target area identification phase, the system estimates the theoretical maximum energy-saving space based on this anchor point to determine whether the current operating condition warrants initiating a speed reduction operation. In the attachment contraction freeze diagram, the region corresponding to the low-confidence freezing marker unit is conservatively increased by 8% in the pre-adjusted speed set to allow for a safety margin in estimation errors. If freezing units in different areas of the attachment contraction freeze diagram appear concentrated in the same time interval, it usually corresponds to the entire sail surface entering a stable attachment state, and the timing sequence of the constant-speed anchor points across the entire pre-adjusted speed set is highly consistent.

[0054] The attachment state constraint correction is implemented based on the pre-adjusted speed set to output the speed adjustment set. The target speed value of the pre-adjusted speed set is read zone by zone. It is then checked whether the lower limit safety boundary of the attachment flow distribution map in each zone remains within the current negative pressure sustainable range when the speed is reduced to the target. Speed ​​targets that do not meet the constraints are gradually increased until they just meet the attachment safety boundary. The corrected set of speed targets for the entire zone constitutes the speed adjustment set. Constraint correction uses the lower limit safety boundary of the attachment flow distribution map in the current frame as the criterion. If the attachment flow distribution map is updated to a new frame during correction, the constraint satisfaction is reassessed using the latest frame. The correction result is always anchored to the latest attachment state boundary. Low-confidence regions with a conservative 8% increase in the pre-adjusted speed set receive an additional increase during the constraint correction phase. Therefore, the final speed target for these low-confidence regions is no lower than the measured speed value required to maintain the attachment boundary. The speed difference between adjacent zones is limited by cooperative constraints: excessive speed difference between adjacent zones along the span can trigger abrupt changes in local pressure gradient, leading to unexpected instability of the attachment state in adjacent zones. Cooperative constraints restrict this transmission chain. When the difference between the corrected constant speed target of a certain zone and that of the adjacent zone exceeds a set range, the target of the adjacent zone in the speed adjustment set is synchronously adjusted upwards to within the difference threshold. Abrupt changes in pressure gradient are particularly sensitive at the mid-section and the boundary between the two ends of the sail. Historically, many attachment instability events have been traced back to insufficient control of speed difference at this location. The cooperative constraint parameters are tuned with reference to the gradient trigger threshold of such historical events. During the transition phase from large to small angle of attack, the constant speed target in most areas of the pre-adjusted speed set shifts significantly downwards under larger angle of attack conditions. During the constraint correction phase, the lower limit of the attachment flow distribution map in each zone shifts synchronously downwards as the angle of attack decreases. The correction conditions are generally met, and the speed adjustment set outputs a significantly reduced constant speed target across the entire zone during this transition condition. The energy-saving effect is immediately reflected in the first adjustment cycle after the speed adjustment set is issued.

[0055] In some embodiments, identifying the lift marginal gain attenuation segment from the speed adjustment set to determine the energy consumption reduction target area includes: performing speed-lift gain characteristic analysis on each region of the speed adjustment set to obtain a gain efficiency table; performing lift saturation plateau region analysis on the gain efficiency table to determine the lift marginal gain attenuation segment; using the lift marginal gain attenuation segment to perform weighted sorting of historical overdrive frequencies to generate an overdrive confidence set; and performing high confidence region identification based on the overdrive confidence set to obtain the energy consumption reduction target area.

[0056] A gain efficiency table is obtained by analyzing the speed-lift gain characteristics of each zone in the speed regulation set. Historical speed-lift response samples around the constant speed target in each zone of the speed regulation set are read, and the marginal gain of lift in each zone at different speed levels is calculated and quantified by the following formula: η(i,n)=ΔL(i,n) / Δn(i), where η(i,n) is the marginal gain of lift in zone i at speed n, with the dimension being lift units / speed units; ΔL(i,n) is the measured increment of lift in zone i when the speed increases from n to Δn(i); Δn(i) is the speed increment step size in zone i, which is determined based on the fan speed resolution of each zone during the initialization phase. The larger η(i,n) is, the higher the lift return per unit speed increase. When η(i,n) approaches zero and no longer increases with speed increase, it indicates that the lift response in that zone has entered the saturation region, and the energy consumption of further speed increase is no longer converted into effective lift. The gain efficiency table is mainly output by the η(i,n) matrix of each zone at each speed level. Historical samples from areas with low-reliability rate markings in concentrated speed regulation zones are weighted to 0.7 times when calculating the gain efficiency table to prevent low-quality rate data from distorting the gain curve shape. For speed ranges with insufficient historical samples and a set minimum value, linear interpolation of adjacent speed range η values ​​is used for estimation. Interpolated entries are marked with interpolation labels in the gain efficiency table. These interpolated speed ranges are not included in the main sequence continuous speed range determination during the lift marginal gain decay segment identification stage; they serve only as auxiliary references. In certain operating conditions, the ship's operating time is relatively short, resulting in scarce samples for corresponding speed ranges. When interpolation labels appear in concentrated areas, it indicates that the reliability of the gain efficiency table for that operating condition segment is low, and a conservative strategy is applied to energy-saving operations for that segment. Under strong inflow conditions, the ship's fans operate at high speeds for extended periods. The corresponding η(i,n) in the gain efficiency table is generally low in the high-speed range, with most areas already at the edge of lift saturation. The identification range of the lift marginal gain decay segment covers most of the suction area in this type of operating condition, thus giving higher priority to the determination of the energy-saving target area.

[0057] The gain efficiency table is analyzed to determine the lift marginal gain attenuation segment by analyzing the lift saturation plateau region. The curve of η(i,n) changing with increasing speed is scanned region by region in the gain efficiency table. A segment where the absolute value of η(i,n) is consistently below a set attenuation threshold for multiple consecutive speed ranges (e.g., at least 3 ranges) and shows no upward trend is identified as the lift marginal gain attenuation segment for that region. The attenuation threshold is taken as the lower quantile of historical effective lift gain; only speed ranges that truly enter the saturation plateau fall within the attenuation segment identification range. The interpolated speed ranges in the gain efficiency table are not included in the main sequence for determining continuous low values. If both sides of an interpolated speed range are measured speeds below the attenuation threshold, the interpolated speed range is allowed to maintain the continuity of the attenuation segment through bridging, but an interpolation bridging label must be added to the corresponding entry for the lift marginal gain attenuation segment. The attenuation segment with interpolated bridging labels applies a slightly conservative confidence level during the overdrive confidence set generation stage, as bridging relies on interpolated values ​​and may virtually connect an originally interrupted attenuation segment into a continuous attenuation region. In the gain efficiency table, a situation where η(i,n) experiences a single abnormally low value at a certain speed range due to turbulent interference does not trigger the identification of the lift marginal gain attenuation segment. An isolated low value must be surrounded by multiple consecutive low values ​​to constitute an effective attenuation segment. The aerodynamic sensitivity of the mid-sail suction region is higher than that of the two ends under most operating conditions. Its η(i,n) curve begins to enter the plateau region at the medium speed range, while attenuation in the two ends usually occurs at higher speed ranges. Therefore, the gain efficiency table shows an uneven attenuation starting point along the span. The spanwise distribution of the lift marginal gain attenuation segment provides a preliminary boundary for the spatial range of the subsequent energy reduction target area. If η(i,n) in the low confidence rate region of the gain efficiency table shows a significantly low value, the continuous gear requirement will be tightened further when identifying the lift marginal gain attenuation section. The falsely low η value caused by calculation error will not be included in the attenuation identification. If the starting speed gear of the lift marginal gain attenuation section in each region shows a monotonically downward trend in cross-batch comparison, it is a signal that the aerodynamic performance of that region is continuously deteriorating with the operating time. The downward shift rate of the attenuation starting point can be used as a quantitative reference for the urgency of sail maintenance.

[0058] For example, the step of using the lift marginal gain attenuation segment to perform weighted sorting of historical overdrive frequencies to generate an overdrive confidence set includes: performing partitioned historical overdrive record analysis on the lift marginal gain attenuation segment to obtain overdrive frequency values ​​for each region; performing time-reverse attenuation weighting processing on the overdrive frequency values ​​for each region to establish a weighted overdrive frequency table; performing frequency sorting for each region based on the weighted overdrive frequency table to generate an overdrive sorting sequence; and performing confidence level mapping based on the overdrive sorting sequence to form an overdrive confidence set.

[0059] Historical overdrive records were analyzed across different regions to obtain the overdrive frequency values ​​for each region, focusing on the lift marginal gain attenuation segment. The archived records of previous overdrive events were retrieved region by region, and overdrive events occurring within the corresponding speed range of the lift marginal gain attenuation segment were selected. After extending the tolerance by half a speed increment at both the upper and lower boundaries of the lift marginal gain attenuation segment, the number of valid overdrive events was counted. Overdrive events within the tolerance range were included with a 0.5x weight, while those outside the tolerance were excluded. The overall statistical results constitute the overdrive frequency values ​​for each region. The tolerance extension is a trade-off: the attenuation segment boundary itself has estimation bias, and without a tolerance, marginal overdrive events would be systematically missed; however, if the tolerance is too wide, overdrive events within the normal speed range would be included in the statistics. The compromise of a half-speed tolerance plus a 0.5x weight ensures that the overdrive frequency values ​​for each region maintain high coverage of overdrive behavior within the actual lift saturation range. When a certain angle of attack condition dominates the flight time, the effective overdrive records corresponding to the lift margin gain decay segment become denser, and the overdrive frequency values ​​in each zone are significantly higher within this condition range. The backtracking window for overdrive records is scaled by the average cycle of the flight mission. This scale ensures that recent sporadic overdrive events are not amplified, nor that long-term events occupy an excessive proportion in the current credibility assessment. The lift margin gain decay segment covers a region with a narrow range of engine speeds, and the total number of effective overdrive events is limited by the range, inevitably resulting in lower overdrive frequency values ​​in each zone. In the weighted overdrive frequency table processing stage, low-frequency regions are supplemented with estimated values ​​from adjacent zones. After the estimated values ​​are marked with low sample values, they are written into the weighted overdrive frequency table, ensuring complete frequency coverage across the entire region. Supplementary estimations do not replace actual statistics. Once the actual sample size in subsequent batches reaches its minimum, the estimated entries are automatically overwritten with measured values, and the overdrive frequency values ​​in each zone continuously converge towards the measured drive mode as the number of flights accumulates.

[0060] A weighted overdrive frequency table is established by applying time-reverse attenuation weighting to the overdrive frequency values ​​of each zone. The occurrence time of the overdrive event corresponding to the overdrive frequency value is extracted for each zone, and an exponential attenuation weight is assigned based on the time interval from the current assessment time. The product of the weight and the event count is accumulated for each entry in the weighted overdrive frequency table for that zone. The attenuation coefficient is calibrated according to the typical degradation period of the sail's aerodynamic characteristics; sails with shorter degradation periods use larger attenuation coefficients to highlight the indicative role of recent conditions. Overdrive events within the tolerance range of each zone's overdrive frequency values, included with a 0.5x weight, inherit this half-weight during time attenuation weighting. After two reductions, the contribution of the corresponding events in the weighted overdrive frequency table is approximately one-quarter of that of events within the formal interval, thus reducing the influence of marginal events on the frequency distribution to a secondary position. In areas where aerodynamic performance has temporarily rebounded due to recent maintenance, the frequency of overdrives has significantly decreased in recent records, while historical records remain high. Time-reverse attenuation weighting dilutes historically high frequencies with recent low frequencies. The corresponding entries in the weighted overdrive frequency table gradually decrease with the accumulation of flights after maintenance. This allows the identification phase of the energy-saving target area to reflect the true improvement trend of overdrive risk in that area, rather than continuously outputting conservative energy-saving judgments based on pre-maintenance high-frequency data. Before selecting the attenuation coefficient, it must be verified through historical batch backtesting. Backtesting must eliminate two mismatches: if the coefficient is too large, a single, occasional high-frequency flight will suddenly increase the value of the entries in the weighted overdrive frequency table, causing a jump in the overdrive ranking sequence; if the coefficient is too small, insufficient dilution of long-term historical data will make it difficult for recent aerodynamic improvements to be reflected in the weighted overdrive frequency table. A coefficient that passes backtesting allows the frequency attenuation curve to closely match the actual evolution of aerodynamic conditions.

[0061] Based on the weighted overdrive frequency table, the overdrive ranking sequence is generated by sorting the frequency of each region from highest to lowest. The entries in the weighted overdrive frequency table are arranged in descending order, with regions having higher weighted frequencies ranking higher. These regions have historically exhibited the strongest inertia in maintaining high speeds within the lift saturation range and are prioritized for energy-saving speed reduction operations. The ranking results for all regions constitute the overdrive ranking sequence. Regions with low-sample annotations in the weighted overdrive frequency table are listed separately as supplementary sequences in descending order and do not participate in the main sequence ranking competition. Regions in the supplementary sequences are subject to a conservative grading strategy when mapping the overdrive confidence set, and their energy-saving operation priority is lower than that of regions of the same level in the main sequence. Adjacent ranking regions with insufficient frequency differences to set a precision threshold are marked with the same level identifier in the overpass sorting sequence. Regions with the same level identifier are grouped into the same confidence level when mapping the overpass confidence set. If frequencies with very similar values ​​swap rankings due to minor fluctuations, the confidence level driven by the overpass sorting sequence will flip back and forth. Merging of the same level eliminates this instability. The precision threshold is selected with reference to the historical standard deviation of the value distribution of all entries in the weighted overpass frequency table. It is appropriately relaxed under the condition of sail surface with large dispersion. The coverage of the same level identifier is matched with the actual frequency discrimination capability. When the operating conditions are highly repetitive and the same intervals are repeatedly experienced, the overpass records of some operating condition intervals are significantly higher than those of other intervals as the number of voyages accumulates. The overpass sorting sequence converges quickly and the ranking is stable. The consistency of the confidence level is better than that of the operating mode with dispersed operating conditions. This statistical advantage allows the weighted overpass frequency table to establish a high-confidence regional frequency benchmark more quickly. The identification accuracy of the energy consumption reduction target area can be improved by about 20% to 30% compared with the early stage.

[0062] An overdrive confidence set is formed by mapping confidence levels based on the overdrive sorting sequence. The absolute value of the weighted frequency of each region in the overdrive sorting sequence is compared with the quantile cut-off point of the frequency distribution across the entire region: regions above the 75th percentile are mapped to a high confidence level, those between the 25th and 75th percentiles are mapped to a medium confidence level, and those below the 25th percentile are mapped to a low confidence level. These three mapping results, along with the region numbers, constitute the overdrive confidence set. The main sequence and supplementary sequences of the overdrive sorting sequence are mapped separately, with the supplementary sequence regions classified according to a conservative classification strategy. The quantile cut-off point is updated in real-time based on the frequency of the entire region at each evaluation, and the confidence level is dynamically adjusted according to the sail status, rather than being fixed to a static threshold. When a region of the same level in the overdrive sorting sequence is near a quantile boundary, the level is uniformly classified based on the level of the average weighted frequency of the same level region, to avoid the quantile cut-off point accidentally cutting the same level region to a different level, thus disrupting the continuous coverage determination of the energy reduction target area. In the first batch of operational data after the overhaul, the values ​​of the weighted overdrive frequency table entries for each zone were distorted due to the partial deletion of historical records, compressing the frequency differences between zones. Some zones that originally had medium frequencies were mapped to higher confidence levels in this batch, resulting in low reliability of the overdrive confidence set classification results. After the system identifies the overhaul flag, it adds an initial overhaul batch label to the overdrive confidence set of this batch. Overdrive confidence sets carrying this label are subject to stricter continuous coverage requirements when identifying energy-saving target areas. The artificially high confidence caused by the scarcity of historical data cannot drive overly aggressive energy-saving operations. High-confidence zones are given priority in the energy-saving target area confirmation stage. Medium-confidence zones are confirmed when there is continuous coverage between adjacent high-confidence zones. Low-confidence zones can only trigger speed reduction operations after the system has sufficient energy-saving margin and the global lift constraint verification is passed.

[0063] Based on the overdrive confidence set, high-confidence regions are identified to obtain the energy reduction target area. Regions in the overdrive confidence set that reach the high confidence level are identified as high-confidence energy reduction candidate regions. When candidate regions are adjacent in the spanwise space and continuously covered, they constitute the entire energy reduction target area. High-confidence records of isolated single regions must meet the continuous coverage condition with adjacent regions to be included in the energy reduction target area. If an individual region experiences a speed reduction due to occasional high-frequency overdrive, a local speed dip will appear in the spanwise direction. The continuous coverage condition excludes this situation. Even if the confidence level of low-sample labeled regions in the overdrive confidence set meets the standard, the confidence level of adjacent regions must still be used as auxiliary evidence when confirming the energy reduction target area. Both must meet the condition to be merged into the continuous energy reduction target area. If the adjacent regions on both sides of a low-sample region are of medium confidence, the low-sample region will not trigger the energy reduction target area identification for the time being. It will be re-evaluated after the sample size in subsequent batches accumulates to the minimum value. During the prolonged stable operation phase after wind speed decreases, the suction system often continues to use the high rotational speed calibrated under the previous high-flow conditions, while the actual lift demand has narrowed as wind speed decreases. This systematic overdrive exhibits a multi-zone synchronous high reliability pattern in the overdrive reliability set. In this scenario, the energy reduction target area covers most of the spanwise region, and active speed reduction has the most significant effect on overall energy consumption reduction. After the energy reduction target area is confirmed, its coverage area number is bound to the overdrive reliability set version number used in this operation. The basis for each energy-saving operation can be traced back to the gain analysis and overdrive history data that generated the results for that batch. If the marginal gain attenuation segment of the lift in the next batch changes significantly, the energy reduction target area driven by the historical version of the overdrive reliability set must be forcibly re-evaluated, and the reliability judgment basis cannot be decoupled from the latest gain characteristics.

[0064] Active speed reduction and overdrive archiving operations are performed in the target area for energy saving to generate energy-saving control quantities. The current constant speed target for each zone within the target area is gradually reduced based on the speed adjustment set according to confidence level step sizes. The reduction step size is larger in high-confidence zones than in medium-confidence zones. The speed reduction magnitude is linked to the reliability of historical overdrive records. Medium-confidence zones do not undergo significant speed reductions until confirmation is completed in high-confidence zones. After each reduction, a stable number of frames is waited to confirm that the lower limit safety boundary of the attached flow distribution map for that zone is not affected. Once confirmed, the reduction amount is recorded as an overdrive archiving entry for that zone. The accumulated overdrive archiving amounts for zones where multiple consecutive reductions have been confirmed constitute a historical available speed reduction space benchmark for reference when generating subsequent batches of energy-saving control quantities. If, during the adjustment process in a certain zone within the energy reduction target area, a boundary contraction acceleration occurs at the corresponding position on the attached flow distribution map, the adjustment in that zone is immediately stopped. Upon stopping, the speed is reset to the previously confirmed speed as the energy-saving control quantity record for that zone in this round, and an early termination marker is added to the overdrive archive entry. The reliability level of the area marked with an early termination marker is not adjusted in the next batch of evaluations, ensuring that a single successful adjustment does not excessively pressure the constant speed target in subsequent batches. An upper limit is set for the overdrive archive quantity in the area marked with an early termination marker. When the cumulative archive quantity exceeds the upper limit, the gain efficiency table for that zone is re-evaluated to determine whether the current lift marginal gain decay segment has shifted due to changes in the sail aerodynamic state. The generation basis for the energy-saving control quantity thus maintains correspondence with the actual sail state. Under operating conditions where lift demand is far below the rated level, the energy reduction target area typically covers most of the spanwise region. Multiple consecutive adjustments in each zone can be confirmed through attachment safety, and the cumulative speed reduction of the energy-saving control quantity is significant under these conditions, resulting in the most outstanding system energy-saving effect.

[0065] The energy-saving control quantities are verified for global lift constraints, and an electronically controlled negative pressure adjustment command is output. After summing the constant speed targets for each zone of the energy-saving control quantities, the estimated lift contribution after speed reduction is accumulated zone by zone, using the minimum safe total lift of the sail under the current operating conditions as the lower limit of the constraint. If the accumulated estimated lift of all zones is still higher than the safe lower limit, the verification passes; if it is lower than the safe lower limit, the speed reduction operation is gradually canceled starting from the zone with the lowest confidence in the overdrive confidence set until the constraint is met. No electronically controlled negative pressure adjustment command is output before the verification passes. The cancellation operation only restores the speed target of the corresponding zone to the original constant speed value of the speed adjustment set, without restarting the under-pumping detection. After the verification passes, the final speed target of the entire zone, together with the corresponding zone negative pressure setpoint, constitutes the electronically controlled negative pressure adjustment command. The negative pressure setpoint is calculated from the constant speed target according to the speed-negative pressure characteristics of each zone and obtained by dynamically calibrating and mapping to the safe upper limit of the current operating conditions. When the operator issues a rapid deceleration command due to maneuvering needs, the energy-saving control quantity is executed in conjunction with this command. The electronically controlled negative pressure adjustment command, while meeting the rapid deceleration requirement, maintains the negative pressure in each zone within the dynamic calibration mapping constraint range. This responds to the control requirements without causing rapid contraction of the attachment boundary due to abrupt deceleration. The speed target of the early abort indicator area in the energy-saving control quantity is calculated with a conservative estimate when verifying the global lift constraint. The conservative estimate is taken as the lower quantile of the lift contribution of the early abort area in similar historical operating conditions. An optimistic lift estimate would mask the actual erosion of the total lift by the energy-saving operation. By using the lower quantile, the electronically controlled negative pressure adjustment command still leaves sufficient lift safety margin in the early abort scenario.

[0066] To implement the above-described method embodiment, a precise negative pressure control method for ship suction sail zones is described to achieve the corresponding functions and technical effects. See [link to relevant documentation]. Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a ship suction sail zoned negative pressure precision control system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The ship suction sail zoned negative pressure precision control system 200 provided in this embodiment includes: Data acquisition module 201 is used to independently acquire pressure data for each suction zone to form a regional pressure difference sequence, and to identify long-term ultra-stable segments of pressure difference from the regional pressure difference sequence to obtain an over-stability anomaly precursor sequence; Risk identification module 202 is used to generate a separation risk index based on the depth weighted amount of the trigger frequency of each region according to the overstability anomaly precursor sequence, screen for long-term pressure stability characteristics based on the separation risk index to determine hidden separation areas, and use the hidden separation areas to track the pressure difference gradient changes between adjacent regions to establish a zonal risk map. The calibration mapping module 203 is used to obtain the partition negative pressure given value by performing a weighted iterative fitting of the sudden change of the incoming flow angle based on the partition risk map, implement the partition negative pressure given value by overload triggering high reliability calibration to construct a self-correcting mapping table, and perform historical overload distribution verification on the self-correcting mapping table to form a dynamic calibration mapping. The collaborative allocation module 204 is used to implement a separation-shielded full-area differential pressure collaborative adjustment output differential pressure adjustment scheme based on the dynamic calibration mapping, and to perform graded under-pumping boundary approximation based on the differential pressure adjustment scheme to obtain the attached flow distribution map. The negative pressure control module 205 is used to perform attachment zone contraction rate tracking and constant speed output speed adjustment set through the attachment flow distribution map, identify the lift marginal gain attenuation segment from the speed adjustment set to determine the energy saving target area, perform active deceleration and overdrive archiving operations for the energy saving target area to generate energy saving control quantity, and perform global lift constraint verification on the energy saving control quantity to output electronically controlled negative pressure adjustment command.

[0067] The aforementioned ship suction sail zoned negative pressure precision control system 200 can implement a ship suction sail zoned negative pressure precision control method according to the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0068] 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.

[0069] 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 precise control of negative pressure in different zones of a ship's suction sail, characterized in that, include: Pressure data is collected independently for each suction zone to form a regional pressure difference sequence. Long-term ultra-stable segments of pressure difference are identified from the regional pressure difference sequence to obtain an abnormal signal sequence of overstability. Based on the statistical analysis of the trigger frequency depth weighting of the overstability anomaly precursor sequence, a separation risk index is generated. The separation risk index is used to screen for long-term pressure stability characteristics to determine hidden separation zones. The hidden separation zones are then used to track the pressure gradient changes between adjacent regions to establish a zonal risk map. Based on the risk map of the partition, a weighted iterative fitting of the sudden change in the angle of attack of the incoming flow is carried out to obtain the negative pressure setpoint of the partition. The negative pressure setpoint of the partition is subjected to overload-triggered high-confidence calibration to construct a self-correcting mapping table. The self-correcting mapping table is verified by historical overload distribution to form a dynamic calibration mapping. Based on the dynamic calibration mapping, a separate shielded full-area differential pressure coordinated regulation output differential pressure regulation scheme is implemented. Based on the differential pressure regulation scheme, a graded under-pumping boundary approximation is performed to obtain the attached flow distribution map. The attachment flow distribution map is used to track the contraction rate of the attachment zone and output a constant speed adjustment set. The lift marginal gain attenuation segment is identified from the speed adjustment set to determine the energy saving target area. Active deceleration and overdrive archiving operations are performed on the energy saving target area to generate energy saving control quantity. The energy saving control quantity is verified by global lift constraint and an electronically controlled negative pressure adjustment command is output.

2. The method according to claim 1, characterized in that, The step of identifying long-term ultra-stable segments of pressure differentials from the regional pressure differential sequence to obtain an over-stability anomaly precursor sequence includes: The pressure difference sequence in the region is subjected to insufficient pulsation intensity quantification to obtain a measure of pressure difference stability; The pressure difference stability measurement is used to screen out the persistently high pressure difference segment and determine the long-term ultra-stable segment of the pressure difference. Based on the aforementioned long-term ultra-stable pressure differential period, a historical ultra-stable frequency comparison and verification were conducted to obtain a stable precursor sequence; Based on the stable precursor sequence, the spread-direction stationary anomaly diffusion region is calibrated to form an overstable anomaly precursor sequence.

3. The method according to claim 1, characterized in that, The process of identifying latent separation regions based on long-term stable pressure characteristics using the separation risk index screening includes: The regional stationarity index is obtained by calculating the time-domain stationarity index of the separation risk index. A long-term continuity test is performed on the aforementioned regional stability index to determine ultra-long stable segments; Based on the aforementioned ultra-long stable segment, multi-region synchronous low-pulsation collaborative identification is implemented to generate a synchronous stable anomaly map; The synchronous stationary anomaly map is used to identify the source of separation and diffusion and determine the hidden separation region.

4. The method according to claim 1, characterized in that, The step of constructing a self-correcting mapping table by implementing overload-triggered high-reliability calibration on the given negative pressure value of the partition includes: Real-time monitoring of the operating status of the fan is performed on the given negative pressure value of the partition to obtain the operating status data. An overload excitation point set is established for the excessive current triggering period of the fan operation status screening. Based on the overload excitation point set, synchronous operating condition parameters are extracted to obtain an overload calibration benchmark library; The overload calibration benchmark library is iteratively weighted and fused to form a self-correcting mapping table.

5. The method according to claim 1, characterized in that, The process of obtaining the attached flow distribution map by performing graded under-pumping boundary approximation based on the differential pressure adjustment scheme includes: Based on the aforementioned differential pressure regulation scheme, the lift redundancy of each zone is quantified to obtain a lift redundancy classification table; Based on the aforementioned lift redundancy classification table, under-pumping timing planning is performed to determine the under-pumping timing table; The under-suction timing table is subjected to critical attachment limit approximation execution to generate partition execution control quantity; The attachment flow distribution map is obtained by using the partition execution control variables to confirm the lower limit safety boundary of the attachment.

6. The method according to claim 1, characterized in that, The method of using the attached flow distribution map to track the contraction rate of the attachment zone and adjust the constant speed output rotation speed includes: The attachment shrinkage rate is obtained by performing temporal frame difference calculation on the attachment flow distribution map. Based on the amount of attachment shrinkage rate, an attachment shrinkage freeze map is obtained by identifying the time-by-time steady segment that approaches zero. Based on the aforementioned attachment shrinkage freezing diagram, a pre-adjusted speed set is established by prioritizing constant speed in the freezing persistence extreme value region. Based on the pre-adjusted speed set, an attachment state constraint correction output speed adjustment set is implemented.

7. The method according to claim 1, characterized in that, The step of identifying the lift marginal gain attenuation segment from the speed regulation set to determine the energy loss reduction target area includes: For the aforementioned speed regulation set, a speed-lift gain characteristic analysis is performed in each zone to obtain a gain efficiency table; The lift efficiency table is analyzed to determine the lift marginal gain attenuation segment by performing lift saturation plateau region analysis. The lift marginal gain attenuation segment is used to perform a weighted sorting of historical overdrive frequencies to generate an overdrive confidence set; Based on the overdrive confidence set, high confidence region identification is performed to obtain the energy reduction target area.

8. The method according to claim 4, characterized in that, The establishment of an overload excitation point set for screening excessive current triggering periods of the fan operating status includes: Multi-phase current decomposition is performed on the fan operating status variables to obtain phase current records; The overload time sequence is obtained by annotating the current plateau segment before overload in the phase current record. Based on the overload time sequence, synchronous operating condition parameters are captured to establish an overload excitation sampling library; Based on the overload excitation sampling library, an excitation point aggregation analysis is performed to establish an overload excitation point set.

9. The method according to claim 7, characterized in that, The process of generating an overdrive confidence set by weighting and sorting historical overdrive frequencies using the lift marginal gain attenuation segment includes: The historical overdrive records of the lift marginal gain attenuation section are analyzed by region to obtain the overdrive frequency value of each region; A weighted overdrive frequency table is established by performing time-reverse attenuation weighting on the overdrive frequency values ​​of each region. Based on the weighted overdrive frequency table, the frequency of each zone is sorted from high to low to generate an overdrive sorting sequence; Based on the overdrive sorting sequence, a confidence level mapping is performed to form an overdrive confidence set.

10. A precise control system for zoned negative pressure of a ship's suction sail, characterized in that, include: The data acquisition module is used to independently acquire pressure data for each suction zone to form a regional pressure difference sequence, and to identify long-term ultra-stable segments of pressure difference from the regional pressure difference sequence to obtain an over-stability anomaly precursor sequence; The risk identification module is used to generate a separation risk index based on the depth weighted average of the trigger frequency of each region according to the overstability anomaly precursor sequence, screen for long-term pressure stability characteristics based on the separation risk index to determine hidden separation areas, and use the hidden separation areas to track the pressure difference gradient changes between adjacent regions to establish a zonal risk map. The calibration mapping module is used to obtain the partition negative pressure given value by performing weighted iterative fitting of sudden change of incoming flow angle based on the partition risk map, implement overload triggered high-confidence calibration of the partition negative pressure given value to construct a self-correcting mapping table, and verify the historical overload distribution of the self-correcting mapping table to form a dynamic calibration mapping. The collaborative allocation module is used to implement a separate shielded full-area differential pressure collaborative adjustment output differential pressure adjustment scheme based on the dynamic calibration mapping, and to perform graded under-pumping boundary approximation based on the differential pressure adjustment scheme to obtain the attached flow distribution map. The negative pressure control module is used to track the contraction rate of the attachment zone and output a constant speed adjustment set through the attachment flow distribution map, identify the lift marginal gain attenuation segment from the speed adjustment set to determine the energy saving target area, perform active deceleration and overdrive archiving operations for the energy saving target area to generate energy saving control quantity, and perform global lift constraint verification on the energy saving control quantity to output an electronically controlled negative pressure adjustment command.

Images