A weather and ice-coating relationship monitoring method based on ice-coating mechanism analysis

Abstract

The application discloses a weather and icing relationship monitoring method based on icing mechanism analysis, and particularly relates to the field of weather and icing relationship monitoring, and comprises the following steps: constructing a factor combination vector sequence containing disturbance amplitude characteristics and icing response synchronous relationship; generating a composite input structure for a dynamic state recognition and determination process by performing weather factor time sequence calculation, disturbance change rate extraction and icing rate synchronous mapping operation, and realizing a disturbance dominant sensitivity state perception data organization mechanism; the application constructs a factor combination vector based on the synchronous mapping mechanism between the disturbance change rate and the icing response, and realizes model boundary self-adaptive adjustment in combination with a nonlinear rate evolution structure, so as to dynamically reflect the fusion trend of state continuous evolution and sudden transition, thereby solving the problem of model response lag in the physical transition of the icing state under complex weather conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorology and icing relationship monitoring, and more particularly to a meteorology and icing relationship monitoring method based on icing mechanism analysis. BACKGROUND

[0002] In the prior art, most icing monitoring methods rely on static meteorological factor threshold judgment or type classification based on empirical rules, lacking dynamic mapping capability of icing rate based on mechanism modeling, resulting in unstable prediction accuracy under complex weather conditions. Even if some methods introduce nonlinear models, they often use fixed decision boundaries, which cannot reflect the continuous evolution of the icing physical process, especially in rain-snow mixed, critical temperature fluctuation and other state transition scenarios, showing decision lag or misjudgment tendency.

[0003] On the basis of existing nonlinear mapping models based on atmospheric factors and mechanism analysis gradually having fine decision-making ability, the system will face a key problem: how to dynamically adjust the model decision-making mechanism when the icing state undergoes physical transition, so as to have the ability to respond to the process continuity and sudden change. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a meteorology and icing relationship monitoring method based on icing mechanism analysis, which constructs a factor combination vector based on the synchronization mapping mechanism between the disturbance change rate and the icing response, and realizes adaptive adjustment of the model boundary by combining the nonlinear rate evolution structure, to dynamically reflect the fusion trend of state continuous evolution and sudden transition, thereby solving the problem of model response lag when the icing state undergoes physical transition under complex meteorological conditions.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a meteorology and icing relationship monitoring method based on icing mechanism analysis, comprising:

[0006] S1, constructing a factor combination vector sequence containing disturbance amplitude characteristics and icing response synchronization relationship, generating a composite input structure for dynamic state recognition and decision-making process by performing meteorological factor time series calculation, disturbance change rate extraction and icing rate synchronization mapping operation, and realizing a sensitivity state perception data organization mechanism under disturbance dominance;

[0007] S2, executing a state transition recognition process based on the disturbance characteristics of the combination vector, analyzing the change trend in the rate prediction sequence and extracting the fast disturbance factor satisfying the threshold judgment rule, outputting a transition classification label for distinguishing between potential transition input and stable input, and realizing the construction of a rate nonlinear response structure driven by disturbance;

[0008] S3, performing a boundary adjustment process, analyzing a factor set structure formed by the combination vector in the disturbance dominant path, judging whether there is a need to adjust the rate response branch of the nonlinear rate mapping model output path through the boundary correction function or the boundary maintenance function, and realizing the adaptive reconstruction mechanism of the boundary behavior under the rate state transition;

[0009] S4, performing a prediction error checking process, performing residual comparison between the rate determination result of each cycle and the ice-coated rate observation data, judging whether there is a persistent rate output deviation behavior under the stable structure, and constructing a systematic deviation identification mechanism based on path consistency and error cause tracking;

[0010] S5, performing a rhythm correction process, performing time scale alignment and rate rhythm consistency comparison on the state trajectory sequence formed by a plurality of cycle combination vectors, identifying the rhythm deviation behavior in the rate output, and constructing a rhythm drift detection mechanism to trigger the dynamic adjustment process of the rhythm correction path.

[0011] In a preferred embodiment, S1 further comprises obtaining atmospheric factor original data recorded in the form of time series by performing a continuous collection operation in the weather and ice relationship monitoring method, and constructing a multi-factor time series set;

[0012] The multi-factor time series set is calculated according to the disturbance amplitude of the fixed time window, and a disturbance intensity data sequence characterized by the disturbance change rate is formed;

[0013] The disturbance intensity data sequence and the ice-coated rate observation data obtained by the weather and ice relationship monitoring method in the historical running period through the same source collection method are executed time synchronization processing, and a response sample set associated with disturbance intensity and ice state transition is constructed;

[0014] The disturbance intensity data in the response sample set consistent with the sampling time of the meteorological observation parameter in the multi-factor time series set is window differential calculated, and the disturbance change rate data is extracted and processed in parallel with the meteorological observation parameter, and a factor combination vector sequence is constructed;

[0015] ;

[0016] Wherein,

[0017] ;

[0018] Wherein is the factor combination vector generated at time ; is the original observation value of the th meteorological factor at time ; is the original observation value of the th meteorological factor in the window the amplitude integral value of the disturbance within; is a time weighting function of the weather factor is a characteristic compression function of the original weather factor value; is a nonlinear standardization mapping function of the disturbance change rate; is a characteristic compression function of the original weather factor value; is the total number of weather factors involved in the calculation; is the length of the disturbance calculation window.

[0019] In a preferred embodiment, S2 further comprises inputting the factor combination vector sequence into a nonlinear icing rate mapping model, performing a rate prediction process, and outputting a continuous-time rate prediction sequence;

[0020] ;

[0021] wherein is the icing rate prediction value at the t-th time point; is the combined disturbance input vector; is the parameter vector of the t-th characteristic mapping kernel; is the total number of kernel mapping paths; is the hyperbolic tangent nonlinear mapping function; is the prediction result normalization function; determines whether the current combination vector in the factor combination vector sequence contains a fast disturbance factor from the disturbance intensity data sequence and extracted according to the disturbance change rate threshold judgment rule, and if so, marks the current combination vector as a potential transition input, otherwise marks the current combination vector as a stable input, and uses the combination vector for static determination branch to perform regular rate output;

[0022] performs state neighborhood interpolation calculation on the combination vector marked as potential transition input and its adjacent combination vectors at the previous and subsequent time points, extracts the time evolution trend and outputs the state change trend vector; marks the combination vector marked as stable input as an interpolation skipped state, and directly inherits the rate output result of the original nonlinear mapping model;

[0023] determines whether the combination vector in the state change trend vector has a continuous rate positive rising feature, and if so, marks the corresponding combination vector as a critical transition input and outputs the combination vector, and if not, marks the corresponding combination vector as a non-transition input and outputs the combination vector as a state stable phase;

[0024]

[0025] ;

[0026] wherein ​​

[0027] ;

[0028] wherein is a label of the current combination vector whether it is a transition input; is a rate prediction result output by the nonlinear model; is a basic threshold for transition determination; is a dynamic trend correction term driven by perturbation; is a perturbation trend weight of the meteorological factor; is an acceleration amount of the meteorological factor .

[0029] In a preferred embodiment, S3 further comprises executing a perturbation direction vector solving operation with the combination vector labeled as critical transition input as input data, outputting an evolution trajectory vector describing the dominant path of the perturbation; skipping the solving step for the combination vector labeled as state stable stage, and maintaining the default output logic in the nonlinear mapping model;

[0030] determining whether the combination vector in the evolution trajectory vector presents a change pattern jointly dominated by multiple atmospheric factors, if yes, extracting the dominant perturbation factors in the combination vector to form a dominant factor set, if no, extracting all the perturbation factors in the combination vector to form a complete factor set.

[0031] In a preferred embodiment, S3 further comprises executing a boundary adjustment determination procedure for the dominant factor set and the complete factor set respectively;

[0032] when the input is the dominant factor set, executing a boundary correction function based on the response structure of the dominant perturbation factor to obtain a boundary adjustment factor; when the input is the complete factor set, executing a boundary maintaining function to obtain a static adjustment factor;

[0033] injecting the dynamic adjustment factor and the static adjustment factor into the output determination link of the nonlinear rate mapping model respectively, corresponding to generate a dynamic correction output path and a static maintaining output path, and finally outputting the rate determination result after boundary response adjustment;

[0034] ;

[0035] wherein is the final rate output after adjustment; is the boundary adjustment factor; is the rhythm correction factor; is the Euclidean norm of the combination perturbation vector; is the gradient of the rate prediction to the input combination vector.

[0036] In a preferred embodiment, S4 further comprises performing error calculation operation on the ice-coating rate observation data obtained in each period and the outputted rate determination result after the boundary response correction in the current period, and outputting a rate prediction residual sequence;

[0037] Performing bias interval judgment on the rate errors in the rate prediction residual sequence to determine whether there is an abnormal value exceeding the preset reasonable range, if there is an abnormal value, performing bias cause backtracking processing on the combination vector corresponding to the abnormal value, and if all errors are within the reasonable range, outputting an error stability label maintaining the current determination state.

[0038] In a preferred embodiment, S4 further comprises taking the combination vector index path identified in the bias cause backtracking processing process as a factor backtracking path, and performing consistency comparison between the combination vector index path contained in the evolutionary trajectory vector;

[0039] If the two paths are consistent, extracting the overlapping section as a systematic offset segment and constructing an error correction index set; if the two paths are inconsistent, marking the factor backtracking path as an isolated error point, and isolating the corresponding combination vector from the feedback adjustment path of the boundary adjustment determination process;

[0040] ;

[0041] Wherein is the residual between the prediction value at the moment and the observation value; is the adjusted rate output; is the real observation rate in the same period; is the error significance threshold; is the index structure of the current path and the main disturbance path; is the index set of the systematic offset point;

[0042] Inputting the error correction index set into the boundary adjustment determination process, performing adjustment function structure updating operation, updating the adjustment function configuration of the dynamic adjustment factor and the static adjustment factor, and outputting a feedback adjustment vector;

[0043] If the current input is marked as an isolated error point in the path consistency comparison result, register the residual information of the combination vector corresponding to the input, and terminate its participation in the boundary adjustment factor updating process; if the current input is not marked as an isolated error point and does not belong to the error correction index set, skip the adjustment operation and maintain the original structure and output state of the model.

[0044] In a preferred embodiment, S5 further includes performing time-scale standardization processing on the periodic combination vectors contained in the continuous multi-period factor combination vector sequence to form time-series state trajectory data under a unified rhythm.

[0045] Determine whether the rate determination result corresponding to each cycle combination vector in the time-series state trajectory data constitutes a periodic offset in the time dimension. If a periodic offset exists, extract the continuous offset segment to form a state drift feature segment. If no periodic offset is formed, output the rhythm maintenance label and maintain the original rate output control path.

[0046] In a preferred embodiment, S5 further includes performing periodic synchronization analysis on the periodic rate offset segments contained in the state drift feature segment and the corresponding adjustment function configuration in the feedback adjustment vector to solve the rhythm drift correction factor; if there is a state drift feature segment but the synchronization matching condition is not met, the offset feature is recorded but the rhythm correction operation is not performed; if there is no state drift feature segment, the correction is skipped and the previous periodic adjustment function configuration remains unchanged.

[0047] The rhythm drift correction factor is used as an additional dynamic adjustment term in the feedback adjustment vector and injected into the disturbance amplitude calculation step in the disturbance intensity data sequence construction process. The sensitivity adjustment parameter of the disturbance change rate is adjusted to generate the disturbance intensity data sequence after rhythm drift correction adjustment.

[0048] ,

[0049] ;

[0050] in This is the current rhythm correction factor; The length of the rhythm cycle; The predicted value is used as a reference rhythm location; The perturbation input vector after the injection of the correction factor; This is an element-wise multiplication operation.

[0051] The technical effects and advantages of this invention are as follows:

[0052] By constructing a rate response state transition mechanism under the perturbation-dominated path, the problem that traditional static boundary models cannot adjust the judgment logic in real time when meteorological conditions change drastically is solved, realizing dynamic perception of the physical continuous evolution process of icing state and adaptive reconstruction of nonlinear prediction boundary, thereby improving the decision-making accuracy of the system in complex meteorological scenarios.

[0053] The disturbance response factor sequence is constructed by a combined vector structure extracted based on meteorological observation factors and disturbance change rate synchronization, and the trend evolution interpolation operation in the state transition determination process is combined to realize sensitive identification of the rate nonlinear change characteristics and structural analysis of the critical transition input, thereby enhancing the response resolution capability of the model to the sudden icing evolution trend.

[0054] By introducing boundary correction and boundary maintenance functions in the model output path, automatically generating dynamic and static adjustment factors according to the dominant factor structure, and performing adaptive reconstruction of the rate prediction path according to the disturbance dominant path change logic, a model output structure that can follow the meteorological driving change and continuously optimize is constructed.

[0055] Through the rhythm drift detection mechanism, the periodic synchronization comparison of the time sequence trajectory of the combined vector is performed, the rhythm correction factor closed-loop adjustment path is constructed, the continuous correction of the multi-period rate output deviation trend is realized, and the structural misjudgment or trend deviation accumulation caused by the periodic environmental rhythm drift is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The method step flowchart of the present application.

[0057] Figure 2 The disturbance structure construction flowchart of the present application.

[0058] Figure 3 The state transition determination flowchart of the present application.

[0059] Figure 4 The boundary adjustment mechanism flowchart of the present application.

[0060] Figure 5 The system error identification and feedback correction flowchart of the present application.

[0061] Figure 6 The rhythm drift correction flowchart of the present application. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0063] With reference to the drawings in the description Figures 1-6 An embodiment of the present application is a meteorological and icing relationship monitoring method based on icing mechanism analysis, comprising:

[0064] S1, construct a factor combination vector sequence containing a disturbance amplitude feature and an icing response synchronization relationship, generate a composite input structure for a dynamic state recognition and decision-making process by performing meteorological factor time series calculation, disturbance change rate extraction, and icing rate synchronization mapping operations, and realize a sensitivity state perception data organization mechanism under disturbance dominance;

[0065] S2, execute a state transition recognition process based on the disturbance characteristics of the combination vector, analyze the change trend in the rate prediction sequence and extract fast disturbance factors that meet the threshold judgment rule, output a transition classification label for distinguishing between potential transition inputs and stable inputs, and realize a rate non-linear response structure construction under disturbance driving;

[0066] S3, execute a boundary adjustment process, analyze the factor set structure composed of combination vectors in the disturbance dominant path, determine whether there is a rate response branch of the non-linear rate mapping model output path that needs to be adjusted by a boundary correction function or a boundary maintenance function, and realize a boundary behavior adaptive reconstruction mechanism under rate state transition;

[0067] S4, execute a prediction error checking process, perform residual comparison between the rate determination result of each cycle and the icing rate observation data, determine whether there is a persistent rate output deviation behavior under stable structure, and construct a systematic deviation identification mechanism based on path consistency and error cause tracking;

[0068] S5, execute a rhythm correction process, perform time scale alignment and rate rhythm consistency comparison on the state trajectory sequence composed of multiple cycle combination vectors, identify the rhythm deviation behavior in the rate output, and construct a rhythm drift detection mechanism to trigger the dynamic adjustment process of the rhythm correction path.

[0069] S1 also includes obtaining atmospheric factor raw data recorded in the form of a time series by performing a continuous collection operation in the meteorological and icing relationship monitoring method, and constructing a multi-factor time series set; the atmospheric factor raw data includes meteorological observation parameters such as environmental temperature, dew point temperature, relative humidity, average wind speed, wind direction angle, precipitation intensity, atmospheric pressure, and solar radiation intensity;

[0070] Perform disturbance amplitude calculation on the multi-factor time series set according to a fixed time window to form a disturbance intensity data sequence characterized by disturbance change rate;

[0071] Perform time synchronization processing on the disturbance intensity data sequence and the icing rate observation data obtained by the same source collection method in the historical operation cycle of the meteorological and icing relationship monitoring method, the icing rate observation data includes the unit time conductor surface icing thickness change rate, the unit length line ice load growth rate, and the incremental icing mass corresponding to the meteorological time point, and construct a response sample set related to disturbance intensity and icing state transition;

[0072] The disturbance intensity data in the response sample set that is consistent with the sampling time of the meteorological observation parameter in the multi-factor time series set is window-differenced to extract disturbance rate data, which is then encoded in parallel with the meteorological observation parameter to construct a factor combination vector sequence for state recognition input;

[0073] ;

[0074] wherein,

[0075] ;

[0076] wherein is the factor combination vector generated at time , and the factor combination vector contains the disturbance rate and original value of each meteorological factor; is the original observation value of the th meteorological factor at time ; is the disturbance amplitude integral value of the meteorological factor within the window ; is the time weighting function of the meteorological factor , and the time weighting function is used to adjust the influence strength of the disturbance at different time points; is a nonlinear normalization mapping function of the disturbance rate; is a feature compression function of the original meteorological factor value; is the total number of meteorological factors participating in the calculation; is the disturbance calculation window length;

[0077] Further, The formula is used to construct the input structure of the nonlinear prediction model, and the core is to solve the time derivative of the original observation value of the atmospheric factor to obtain the disturbance trend, and then form a dynamic description with "disturbance amplitude memory" through time weighting integration. After the disturbance rate part and the original factor value are input in parallel, two nonlinear encoding functions are used for unified scale conversion to ensure that the input vector has high-dimensional disturbance response capability, which facilitates subsequent nonlinear mapping model identification of rate mode changes.

[0078] S2 further includes inputting the factor combination vector sequence into a nonlinear icing rate mapping model to perform rate prediction processing and output a continuous time rate prediction sequence;

[0079] ;

[0080] wherein is the icing rate prediction value at the th time; is the combined disturbance input vector; is the parameter vector of the th feature mapping kernel; is the total number of kernel mapping paths; is the hyperbolic tangent nonlinear mapping function; is the prediction result normalization function;

[0081] Further, is the main structure of the nonlinear mapping model, which combines the disturbance group vector into the rate output using the activation function to construct the nonlinear mapping of each kernel channel, using the multi-channel structure to perform generalized feature extraction on the input disturbance dimension, and then compressing the output to a physically reasonable interval through the normalization function , which allows the model to adapt to multi-mode disturbance characteristics, nonlinear amplitude regulation, and reserves a clear main path for subsequent adjustment modules;

[0082] determines whether the current combination vector in the combination vector sequence contains the fast disturbance factor extracted from the disturbance intensity data sequence and according to the disturbance rate threshold judgment rule, if it contains, the current combination vector is marked as potential transition input, otherwise the current combination vector is marked as stable input, and the combination vector is used for static judgment branch to execute the regular rate output;

[0083] performs state neighborhood interpolation calculation on the combination vector marked as potential transition input and its adjacent combination vectors at the previous and next time points, extracts the time evolution trend and outputs the state change trend vector; the combination vector marked as stable input is marked as interpolation skip state, and directly inherits the rate output result of the original nonlinear mapping model;

[0084] determines whether there is a continuous rate positive rising feature in the combination vector in the state change trend vector, if there is, the corresponding combination vector is marked as critical transition input and output, if there is not, the corresponding combination vector is marked as non-transition input and output as the state stable stage;

[0085] ;

[0086] wherein

[0087] ;

[0088] wherein is the label of whether the current combination vector is a transition input, 1 for transition input, 0 for stable input; a rate prediction result output by the nonlinear model; a basic threshold for transition determination; a dynamic trend correction term driven by the disturbance; a disturbance trend weight of the mth meteorological factor; an acceleration amount of the meteorological factor; an acceleration amount of the meteorological factor; an acceleration amount of the meteorological factor;

[0089] Further, the formula is used to identify whether the rate prediction corresponding to the combination vector has a transition tendency, the second-order time derivative based on the rate output represents the nonlinear "acceleration" trend, the dynamic correction term is a high-order structure description of the disturbance trend, the acceleration amount of each atmospheric factor is combined and weighted and summed to form an adaptive transition threshold that is adjusted in real time, and the disturbance-sensitive transition determination logic is introduced, which is a key source of model robustness and interpretability.

[0090] S3 further includes inputting the combination vector marked as a critical transition into the nonlinear mapping model as input data, performing a disturbance direction vector solving operation, and outputting an evolution trajectory vector describing the dominant path of the disturbance; and skipping the solving step for the combination vector marked as a state stable stage and maintaining the default output logic in the nonlinear mapping model;

[0091] determining whether the combination vector in the evolution trajectory vector presents a change mode jointly dominated by multiple atmospheric factors, if the combination vector presents the change mode jointly dominated by multiple atmospheric factors, extracting the dominant disturbance factors in the combination vector to form a dominant factor set, and if the combination vector does not present the change mode jointly dominated by multiple atmospheric factors, extracting all disturbance factors in the combination vector to form a complete factor set for subsequent processing.

[0092] S3 further includes inputting the dominant factor set and the complete factor set as independent inputs, respectively, and sequentially executing a boundary adjustment determination process;

[0093] when the input is the dominant factor set, executing a boundary correction function based on the response structure of the dominant disturbance factor to obtain a boundary adjustment factor used for dynamically adjusting the rate output; and when the input is the complete factor set, executing a boundary maintenance function to obtain a static adjustment factor for maintaining the existing output path;

[0094] injecting the dynamic adjustment factor and the static adjustment factor into the output determination link of the nonlinear rate mapping model, respectively, to correspondingly generate a dynamic correction output path and a static maintenance output path, and finally outputting a rate determination result after boundary response adjustment as the icing state evaluation basis for the current period;

[0095] ;

[0096] wherein is the adjusted final rate output; is the boundary adjustment factor, the boundary adjustment factor is used to enhance the perturbation response; is the rhythm correction factor, the rhythm correction factor is used to correct the long-term rhythm deviation; is the Euclidean norm of the combined perturbation vector, the Euclidean norm of the combined perturbation vector represents the overall perturbation intensity; is the gradient of the rate prediction on the input combined vector;

[0097] Further, The formula is used to act the "boundary adjustment factor" and the "rhythm correction factor" on the main model output path, dynamically adjusting the prediction result , the construction logic includes two adjustment terms, one is the log nonlinear response based on the combined vector intensity norm, the nonlinear response is controlled by , and the other is the enhanced response of the rate output to the input direction gradient, the enhanced response is controlled by , the essence of this structure is to establish a coupling mechanism among the "perturbation rhythm response", which improves the adaptability of the model to long-term rhythm and short-term boundary fluctuations.

[0098] S4 further comprises performing error calculation operations on the ice-coated rate observation data obtained in each period and the rate determination result output in the current period after the boundary response correction, and outputting a rate prediction residual sequence;

[0099] Performing deviation interval judgment on the rate error in the rate prediction residual sequence to determine whether there is an abnormal value that exceeds the preset reasonable range, if there is an abnormal value, performing deviation cause backtracking processing on the combined vector corresponding to the abnormal value, and if all errors are within the reasonable range, outputting an error stability label that maintains the current determination state.

[0100] S4 further comprises taking the combined vector index path identified in the deviation cause backtracking processing process as a factor backtracking path, and performing consistency comparison between the combined vector index path contained in the evolution trajectory vector;

[0101] If the two paths are consistent, extract the overlapping section as a systematic deviation segment and construct an error correction index set; if the two paths are inconsistent, mark the factor backtracking path as an isolated error point, and isolate the corresponding combined vector from the feedback adjustment path of the boundary adjustment determination process;

[0102] ;

[0103] wherein is the residual between the prediction value at the moment and the observation value; is an adjusted rate output; is a same-period true observation rate; is an error significance threshold; is an index structure of the current path and the main disturbance path; is an index set of points determined to be systematic drift points;

[0104] Further, is a residual error of the input through the residual error The calculation of the path consistency determination combined with the path consistency determination identifies structural drift samples, and the path represents and the reference path The consistency comparison of the path and the reference path ensures that the error is not caused by an occasional or local disturbance, but is caused by a model structure drift. The construction method of this path error coupling is a key innovation point of system error identification, which can trigger directional adjustment path update or error isolation.

[0105] The error correction index set is input into the boundary adjustment determination process, the adjustment function structure update operation is performed, the adjustment function configuration of the dynamic adjustment factor and the static adjustment factor is updated, and the feedback adjustment vector is output.

[0106] If the current input is marked as an isolated error point in the path consistency comparison result, the residual error information of the combined vector corresponding to the input is recorded, and the input is terminated from participating in the boundary adjustment factor update process. If the current input is not marked as an isolated error point and does not belong to the error correction index set, the adjustment operation is skipped and the original structure and output state of the model are maintained.

[0107] S5 further includes performing time scale standardization processing on each period combined vector contained in the factor combined vector sequence of a plurality of consecutive periods to form time series state trajectory data under a unified rhythm;

[0108] It is judged whether the rate determination result corresponding to each period combined vector in the time series state trajectory data constitutes a periodic drift in the time dimension. If there is a periodic drift, the state drift feature segment is extracted from the continuous drift section. If there is no periodic drift, a rhythm retention label is output, and the original rate output control path is maintained.

[0109] S5 further includes performing periodic synchronization analysis on each period rate drift section contained in the state drift feature segment and the corresponding adjustment function configuration in the feedback adjustment vector to solve the rhythm drift correction factor. If there is a state drift feature segment but does not meet the synchronization matching condition, the drift feature is recorded but the rhythm correction operation is not performed. If there is no state drift feature segment, the correction is skipped and the adjustment function configuration of the previous period is maintained unchanged.

[0110] The rhythm shift correction factor is injected into the disturbance amplitude calculation step in the disturbance intensity data sequence construction process as an additional dynamic adjustment item in the feedback adjustment vector, the sensitivity adjustment parameter of the disturbance change rate is adjusted, and the disturbance intensity data sequence adjusted by the rhythm shift correction adjustment is generated as the closed-loop input for subsequent state disturbance identification, boundary adjustment processing and rhythm correction path update.

[0111] ,

[0112] ;

[0113] wherein is the current rhythm correction factor; is the rhythm period length; is the predicted value at the reference rhythm position; is the disturbance input vector after the injection of the correction factor; is the element-by-element multiplication operation, which means multiplying the vector components at the corresponding positions to form a vector result of the same dimension;

[0114] Further, in In the formula, the current rate prediction sequence is subtracted from the corresponding point of the previous period to obtain the translation amount The difference between the two is minimized to align the rhythm positions, and the optimization target is built on the continuity of the rate, which has a period closed-loop logic basis. Subsequently, the correction factor is injected into the original combination vector to form a disturbance enhancement input. This rhythm synchronization driven data reconstruction path has a dynamic adjustment path with cross-period prediction balance capability, and constitutes the core calculation model of the rhythm feedback mechanism.

[0115] It should be noted that for the formula structure involved in the present scheme, the dimensionless item can be used as a proportional or structural adjustment factor. When combined with quantities with units, it only plays a numerical scaling role and does not introduce new physical dimensions, so it will not change or confuse the unit system of the overall expression. Such a combination of "dimensionless items and quantity unit items" can be understood as a composite structure expression commonly used in mathematical and physical modeling, which conforms to the principle of dimensional consistency and has a clear physical interpretation basis.

[0116] Secondly, in the formula structure of the present scheme, if multiple variables with different physical units are involved, including but not limited to time, mass or energy variables, their joint occurrence is to express the cooperative modeling relationship of multiple physical mechanisms. Each variable can be mapped by a function, combined by a ratio, or adjusted by a normalization to form a unified structure. The unit is clear and the meaning is clear. The overall expression conforms to the principle of dimensional consistency and the common norm of engineering modeling.

[0117] If the design constant, weight, adjustment factor, threshold parameter, proportion coefficient and the like in the scheme are adjusted, they are the adjustable control parameters for different application environments, and their values depend on the target device configuration, data input characteristics and performance optimization target, and are set within a reasonable range through model verification, performance constraints or engineering calibration in the implementation stage. Although such parameters do not have a unique value, they have a clear adjustment logic and calculation path, and belong to the deterministic setting process in engineering implementation. The purpose of such setting is to ensure that the scheme has both general adaptability and reproducibility and operability, and does not affect its technical clarity and implementability.

[0118] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A weather and ice-coating relationship monitoring method based on ice-coating mechanism analysis, characterized in that, Comprise: S1, construct a factor combination vector sequence containing the disturbance amplitude characteristic and the icing response synchronization relationship, generate a composite input structure for the dynamic state recognition decision process by performing meteorological factor time series calculation, disturbance change rate extraction and icing rate synchronization mapping operation, realize the sensitivity state perception data organization mechanism under the disturbance dominance; S2, execute the state transition recognition process based on the disturbance characteristics of the combined vector, analyze the change trend in the rate prediction sequence and extract the fast disturbance factor that meets the threshold judgment rule, output the transition classification label for distinguishing between potential transition input and stable input, and realize the construction of the rate nonlinear response structure driven by disturbance; S3, execute the boundary adjustment process, analyze the factor set structure composed of the combined vector in the disturbance dominant path, judge whether there is a rate response branch of the nonlinear rate mapping model output path that needs to be adjusted by the boundary correction function or the boundary maintenance function, realize the adaptive reconstruction mechanism of boundary behavior under rate state transition; S4, execute the prediction error checking process, compare the rate determination result of each cycle with the icing rate observation data, judge whether there is a persistent rate output deviation behavior under stable structure, and construct a systematic deviation identification mechanism based on path consistency and error cause tracking; S5, execute the rhythm correction process, perform time scale alignment and rate rhythm consistency comparison on the state trajectory sequence composed of multiple cycle combined vectors, identify the rhythm deviation behavior in the rate output, and construct a rhythm drift detection mechanism to trigger the dynamic adjustment process of the rhythm correction path.

2. The weather and icing relationship monitoring method based on icing mechanism analysis according to claim 1, characterized in that: S1 further comprises obtaining the original data of atmospheric factors recorded in the form of time series by performing continuous collection operation in the meteorological and icing relationship monitoring method, and constructing a multi-factor time series set; Perform disturbance amplitude calculation on the multi-factor time series set according to fixed time window to form disturbance intensity data sequence characterized by disturbance change rate; Perform time synchronization processing on the disturbance intensity data sequence and the icing rate observation data obtained by the same source collection method in the historical running period of the meteorological and icing relationship monitoring method, and construct a response sample set related to disturbance intensity and icing state transition; Perform window difference calculation on the disturbance intensity data in the response sample set that is consistent with the sampling time of the meteorological observation parameters in the multi-factor time series set, and perform parallel encoding processing on the disturbance change rate data after extraction and the meteorological observation parameters, to construct a factor combination vector sequence; ； Wherein, ； wherein is the factor combination vector generated at time ; is the original observation value of the th meteorological factor at time ; is the disturbance amplitude integral value of the meteorological factor over the window ; is the time weighting function of the meteorological factor ; is the nonlinear normalization mapping function of the disturbance change rate; is the feature compression function of the original meteorological factor value; is the total number of meteorological factors involved in the calculation; is the disturbance calculation window length.

3. The weather and icing relationship monitoring method based on icing mechanism analysis according to claim 2, characterized in that: S2 further comprises inputting the factor combination vector sequence into the nonlinear icing rate mapping model to perform rate prediction processing, and outputting the rate prediction sequence at consecutive time; ； wherein is the ice accretion rate prediction at time is the combined disturbance input vector; is the parameter vector of the is the total number of kernel mapping paths; is the hyperbolic tangent nonlinear mapping function; is the prediction result normalization function; Determine whether the current combined vector in the factor combination vector sequence contains the fast disturbance factor extracted from the disturbance intensity data sequence according to the disturbance change rate threshold judgment rule, if yes, mark the current combined vector as potential transition input, otherwise mark the current combined vector as stable input, and use the combined vector for static decision branch to perform regular rate output; The state neighborhood interpolation calculation is performed on the combination vector marked as potential transition input and its adjacent combination vectors at the previous and next continuous time points, the time evolution trend is extracted, and a state change trend vector is output; the combination vector corresponding to the marked stable input is marked as an interpolation skip state, and the rate output result of the original nonlinear mapping model is directly inherited; It is judged whether the combination vector in the state change trend vector has a continuous rate positive rising feature, if so, the corresponding combination vector is marked as a critical transition input and output, if not, the corresponding combination vector is marked as a non-transition input and output as a state stable stage; ； Wherein ； wherein is a label for the current combination vector whether it is a transition input or not; is a rate prediction result output by the nonlinear model; is a base threshold for transition determination; is a disturbance driven dynamic trend correction term; is a disturbance trend weight for the th meteorological factor; is an acceleration amount of the meteorological factor .

4. The weather and icing relationship monitoring method based on icing mechanism analysis according to claim 3, characterized in that: S3 further comprises executing a disturbance direction vector solving operation on the combination vector marked as a critical transition input as input data, and outputting an evolution trajectory vector describing the evolution trajectory of the dominant path of the disturbance; the combination vector marked as a state stable stage is skipped in the solving step, and the default output logic in the nonlinear mapping model is maintained; It is judged whether the combination vector in the evolution trajectory vector presents a change mode jointly dominated by multiple atmospheric factors, if it contains a change mode jointly dominated by multiple atmospheric factors, the dominant disturbance factors in the combination vector are extracted to form a dominant factor set, if it does not contain a change mode jointly dominated by multiple atmospheric factors, all disturbance factors in the combination vector are extracted to form a complete factor set.

5. The weather and icing relationship monitoring method based on icing mechanism analysis according to claim 4, characterized in that: S3 further comprises executing a boundary adjustment judgment process on the dominant factor set and the complete factor set as independent inputs in turn; When the input is the dominant factor set, a boundary correction function is executed based on the response structure of the dominant disturbance factor to obtain a boundary adjustment factor; When the input is the complete factor set, a boundary maintenance function is executed to obtain a static adjustment factor; The dynamic adjustment factor and the static adjustment factor are respectively injected into the output judgment link of the nonlinear rate mapping model, corresponding to generate a dynamic correction output path and a static retention output path, and finally output the rate judgment result after boundary response adjustment; ； wherein is the adjusted final rate output; is the boundary adjustment factor; is the rhythm correction factor; is the Euclidean norm of the combined perturbation vector; is the gradient of the rate prediction on the input combined vector.

6. The weather and ice relationship monitoring method based on ice mechanism analysis according to claim 5, characterized in that: S4 further comprises executing an error calculation operation on the ice shedding rate observation data obtained in each period and the rate judgment result output in the current period after boundary response correction, and outputting a rate prediction residual sequence; The rate error in the rate prediction residual sequence is subjected to a deviation interval judgment to determine whether there is an abnormal value exceeding the preset reasonable range, if there is an abnormal value, the deviation cause of the abnormal corresponding combination vector is processed, and if all errors are within the reasonable range, an error stable label maintaining the current judgment state is output.

7. The weather and ice relationship monitoring method based on ice mechanism analysis according to claim 6, characterized in that: S4 further comprises taking the combination vector index path identified in the deviation cause backtracking process as a factor backtracking path, and performing consistency comparison with the combination vector index path contained in the evolution trajectory vector; If the two paths are consistent, the overlapping section is extracted as a systematic deviation segment and an error correction index set is constructed; if the two paths are inconsistent, the factor backtracking path is marked as an isolated error point, and the corresponding combination vector is isolated from the feedback adjustment path of the boundary adjustment judgment process. ； wherein is the first is the residual between the forecast and the observed value at time t; is the adjusted rate output; is the true observed rate for the same period; is the error significance threshold; is the index structure of the current path and the main disturbance path; is the index set of the points determined as systematic drift points; The error correction index set is input into a boundary adjustment judgment process, a regulation function structure update operation is performed, the regulation function configuration of the dynamic adjustment factor and the static adjustment factor is updated, and a feedback adjustment vector is output; If the current input is marked as an isolated error point in the path consistency comparison result, the residual information of the input corresponding combination vector is registered, and the input is terminated from participating in the boundary adjustment factor update process; if the current input is not marked as an isolated error point and does not belong to the error correction index set, the adjustment operation is skipped and the original structure and output state of the model are maintained.

8. The weather and ice relationship monitoring method based on ice mechanism analysis according to claim 7, characterized in that: S5 further includes performing time scale standardization processing on each period combination vector contained in the factor combination vector sequence of the continuous multiple periods to form time series state trajectory data under a unified rhythm; It is judged whether the rate determination result corresponding to each period combination vector in the time series state trajectory data constitutes a periodic shift in the time dimension, if there is a periodic shift, the continuous shift section is extracted to form a state drift feature section, if there is no periodic shift, a rhythm retention label is output, and the original rate output control path is maintained.

9. The weather and ice relationship monitoring method based on ice mechanism analysis of claim 8, wherein: S5 further includes performing periodic synchronization analysis on each period rate shift section contained in the state drift feature section and the corresponding regulation function configuration in the feedback adjustment vector to solve the rhythm drift correction factor; if there is a state drift feature section but does not meet the synchronization matching condition, the offset feature is recorded but the rhythm correction operation is not performed, and if there is no state drift feature section, the correction is skipped and the regulation function configuration of the previous period is maintained unchanged; The rhythm drift correction factor is used as an additional dynamic adjustment item in the feedback adjustment vector, the disturbance amplitude calculation step in the disturbance intensity data sequence construction process, the sensitivity adjustment parameter of the disturbance change rate is adjusted, and the disturbance intensity data sequence after rhythm drift correction adjustment is generated; ， ； wherein is a current rhythm correction factor; is a rhythm period length; is a predicted value at a reference rhythm position; is the perturbation input vector after injection of the correction factor; is an element-wise multiplication operation.

Images