Observer-based disturbance estimation method, apparatus and device, and medium

Abstract

The invention provides a disturbance estimation method and device based on an observer, equipment and a medium, is used for the technical field of automatic control, and can solve the technical problem of low disturbance estimation accuracy caused by switching impact of the observer in the prior art. The method comprises the following steps: determining observation index values respectively corresponding to a high-order observer and a low-order observer according to corresponding state vectors, and judging whether a target observation index value exists in the observation index values or not; when the judgment result is that the target observation index value does not exist, calculating a high-frequency energy ratio according to the energy values corresponding to the target frequency band and the high-frequency frequency band respectively; substituting the high-frequency energy ratio into the target weight model to obtain a corresponding dynamic low-order weight and a dynamic high-order weight; when the judgment result is that the target observation index value exists, determining a preset low-order weight and a preset high-order weight; performing weighted summation on the disturbance estimated value according to the low-order weight and the high-order weight to obtain a target disturbance estimated value; therefore, the accuracy of disturbance estimation is improved.

Description

Technical Field

[0001] This application relates to the field of automatic control technology, and in particular to a disturbance estimation method, apparatus, device and medium based on an observer. Background Technology

[0002] Highly dynamic moving bodies will face unavoidable disturbances during flight. Extended state observers, as an effective disturbance estimation tool, can expand the total disturbance of the system into a new state variable and perform real-time disturbance estimation through the observer. This allows for subsequent feedforward compensation based on the estimated disturbance value, thereby improving the robustness of the system. The order of traditional state observers is fixed in the design stage. Low-order observers have strong noise resistance and are suitable for stable flight, i.e., single-peak disturbance scenarios, while high-order observers have strong tracking capabilities and are suitable for complex flight, i.e., multi-peak disturbance scenarios.

[0003] Existing disturbance estimation methods typically switch observers according to the actual scenario to estimate the corresponding disturbance value. This is done by adjusting the observer gain based on flight parameters (such as Mach number and angle of attack) in a feedforward manner, or by monitoring the statistical characteristics of the observer residual (output estimation error), such as amplitude and root mean square value, to trigger the structure switching and thus achieve the observer switching.

[0004] However, existing disturbance estimation methods suffer from large estimation errors. Summary of the Invention

[0005] This application provides an observer-based disturbance estimation method, apparatus, device, and medium to address the problem of large estimation errors in existing disturbance estimation methods.

[0006] Firstly, this application provides an observer-based perturbation estimation method, the method comprising: Based on the corresponding state vector, determine the observation index values ​​corresponding to the high-order observer and the low-order observer respectively, and based on their respective target thresholds, determine whether there is a target observation index value among the observation index values, and obtain the corresponding judgment result; the target observation index value is the observation index value that exceeds the target threshold. When the judgment result is that there is no target observation index value, the corresponding spectrum signal is determined, and the energy ratio is calculated based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, to obtain the high frequency energy ratio; the high frequency energy ratio is substituted into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high frequency band are the signal frequency bands intercepted from the spectrum signal; When the judgment result indicates that a target observation index value exists, the preset low-order weight and preset high-order weight are determined according to the target observer category corresponding to the target observation index value. Based on the corresponding low-order and high-order weights, the perturbation estimates output by the low-order and high-order observers are weighted and summed to obtain the target perturbation estimate.

[0007] In some embodiments of this application, the observation index values ​​corresponding to the higher-order observer and the lower-order observer are determined based on the corresponding state vectors, including: Based on the state vectors of the higher-order and lower-order observers in the current control cycle, the corresponding output estimates are determined respectively. Based on the roll rate measured by the gyroscope, the difference between the roll rate value and the output estimate is calculated to obtain the higher-order residual signal and the lower-order residual signal. Calculate the matrix traces of the estimation error covariance matrices corresponding to the higher-order and lower-order observers respectively to obtain the higher-order matrix traces and lower-order matrix traces; Determine the residual signal amplitude corresponding to each residual signal, and based on the corresponding residual signal amplitude and matrix trace, determine the observation index values ​​corresponding to the higher-order observer and the lower-order observer.

[0008] In some embodiments of this application, based on their respective target thresholds, it is determined whether a target observation index value exists among the observed index values, and the corresponding determination result is obtained, including: Based on the residual signal amplitude and matrix trace value corresponding to each observed index value, the corresponding comparison results are obtained by comparing each residual signal amplitude and matrix trace value with their respective target thresholds. If the comparison result shows that neither the residual signal amplitude nor the matrix trace value exceeds the corresponding target threshold, then the corresponding judgment result is determined to be that there is no target observation index value. If at least one of the residual signal amplitude and matrix trace exceeds the corresponding target threshold, the corresponding target observer category is determined, and the judgment result is determined to be that a target observation index value exists.

[0009] In some embodiments of this application, determining the corresponding spectral signal includes: Based on any residual signal, the signal within the sliding window is updated to obtain the latest window signal, and a fast Fourier transform is performed on the latest window signal to obtain the spectrum signal.

[0010] In some embodiments of this application, before substituting the high-frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weights and dynamic high-order weights, the method further includes: The hyperbolic tangent model is determined as the model framework corresponding to the target weight model, the initial weight model is obtained, and the corresponding perturbation signal is determined based on the simulation environment; Based on the disturbance signal, the disturbance estimate of the output of the initial weight model is determined, and the energy ratio parameter threshold and control parameters in the initial weight model are optimized according to the disturbance estimate to obtain the target weight model.

[0011] In some embodiments of this application, based on the disturbance estimation value, the energy ratio parameter threshold and control parameters in the initial weighting model are optimized to obtain the target weighting model, including: Based on the simulation environment, the target disturbance value corresponding to the disturbance signal is determined, and the corresponding loss value is determined according to the loss function between the target disturbance value and the disturbance estimate. Based on the loss value and the preset loss threshold, the energy ratio parameter threshold and control parameters are optimized so that the optimized loss value is less than the loss threshold. The target weight model is determined based on the optimized energy ratio parameter threshold and control parameters.

[0012] In some embodiments of this application, preset low-order weights and preset high-order weights are determined based on the target observer category corresponding to the target observation index value, including: If the target observer category is a low-order observer, then the preset low-order weight is set to 0 and the preset high-order weight is set to 1. If the target observer category is a high-order observer, then the preset low-order weight is set to 1 and the preset high-order weight is set to 0.

[0013] Secondly, this application provides an observer-based disturbance estimation device, the device comprising: The judgment module is used to determine the observation index values ​​corresponding to the high-order observer and the low-order observer respectively based on the corresponding state vector, and to determine whether there is a target observation index value among the observation index values ​​according to their respective target thresholds, and to obtain the corresponding judgment result; the target observation index value is the observation index value that exceeds the target threshold. The calculation module is used to determine the corresponding spectrum signal when the judgment result is that there is no target observation index value, and to calculate the energy ratio based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, so as to obtain the high frequency energy ratio; and to substitute the high frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high frequency band are the signal frequency bands extracted from the spectrum signal; The determination module is used to determine the preset low-order weights and preset high-order weights based on the target observer category corresponding to the target observation index value when the judgment result indicates that a target observation index value exists. The summation module is used to perform a weighted summation of the perturbation estimates output by the low-order and high-order observers, respectively, based on the current corresponding low-order and high-order weights, to obtain the target perturbation estimate.

[0014] Thirdly, this application provides a computer device, including: a processor, and a memory communicatively connected to the processor; The memory stores instructions that the computer executes; The processor executes computer execution instructions stored in memory to implement the method of this application.

[0015] Fourthly, this application provides a computer-readable storage medium storing program code, which, when executed by a processor, is used to implement the method of this application.

[0016] Compared with existing technologies, the method in this application is based on the judgment of the observed index value and the target threshold, thereby monitoring the operating status of high-order and low-order observers in real time, accurately identifying and quickly isolating abnormal observers, effectively avoiding the pollution of the final disturbance estimation by the distorted data of the faulty observer, and ensuring the reliability of control decisions under extreme conditions. When all observers are operating normally, the high-frequency energy ratio is obtained by extracting the spectrum signal and calculating the energy ratio of the target frequency band and the high-frequency band. This achieves lightweight and high-sensitivity identification of structural changes in the disturbance spectrum from "single-peak concentration" to "multi-peak dispersion", solving the problem of false triggering caused by sensor interference in traditional residual monitoring. The high-frequency energy ratio is substituted into the target weight model to obtain dynamic low-order weights and dynamic high-order weights. Through continuous and smooth weight mapping, the low-noise of the low-order observer is reduced. The acoustic advantage is fully utilized under stable operating conditions, and the strong tracking capability of high-order observers is efficiently adapted to complex disturbance conditions, solving the transient impact caused by traditional hard switching of order and ensuring the continuity of disturbance estimation output. When abnormal observers are present, preset weights are matched according to the observer category to ensure that the system quickly switches to the disturbance estimation mode dominated by normal observers, avoiding fault propagation that could lead to system failure. By using low-order and high-order weights corresponding to different scenarios in real time, the disturbance estimation values ​​of the two types of observers are weighted and summed to obtain target disturbance estimation values ​​that combine high accuracy and high robustness. This improves the attitude and trajectory control stability of high-dynamic moving bodies under extreme conditions such as transonic flight, while significantly reducing the computational load, improving the scenario adaptability and estimation accuracy of disturbance estimation, and enhancing the estimation precision. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] Figure 1 A flowchart illustrating an observer-based disturbance estimation method provided in this application embodiment; Figure 2 A schematic diagram of the architecture of an observer-based perturbation estimation method provided in an embodiment of this application; Figure 3 A schematic diagram of an observer-based disturbance estimation method provided in this application embodiment; Figure 4 A schematic diagram of the calculation process of an observer-based disturbance estimation method provided in an embodiment of this application; Figure 5 A schematic diagram of an observer-based disturbance estimation method provided in an embodiment of this application; Figure 6 A schematic diagram of the structure of an observer-based disturbance estimation device provided in this application embodiment; Figure 7 This is a structural block diagram of an apparatus for performing an observer-based perturbation estimation method according to an embodiment of this application. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0020] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0021] In practical scenarios, the order of a traditional ESO (Extended State Observer) is fixed during the design phase; to cover potential high-frequency disturbance components, a higher order (such as fourth order) is typically chosen in the design. However, during the stable flight phase where the disturbance energy is concentrated at a single dominant frequency, a higher-order ESO, due to its wider bandwidth, introduces unnecessary measurement noise, thus reducing estimation accuracy. Conversely, if a lower order, such as second order, is chosen for optimization during the stable phase, its limited bandwidth and dynamic response capability will be unable to simultaneously track multiple frequency components when the disturbance spectrum abruptly becomes multi-peaked. This leads to estimation lag, amplitude attenuation, and ultimately, severe phase mismatch, causing compensation based on this distorted estimate to actually worsen system stability.

[0022] To address time-varying disturbances, the gain scheduling engine (ESO) adjusts the observer gain in a feedforward manner based on flight parameters (such as Mach number and angle of attack). However, in the transonic range, the abrupt change window of the disturbance spectrum is extremely short, and the mapping relationship between the abrupt change point and flight parameters is significantly dispersed due to uncertainties such as individual ballistic differences and manufacturing tolerances. Furthermore, the non-negligible delays in sensor measurement, processing, and transmission make it difficult for the feedforward scheduling strategy to accurately match the actual disturbance structure changes, resulting in an inherent "scheduling lag" problem.

[0023] Variable-order / variable-structure ESO based on residual monitoring: This method triggers structure switching by monitoring the statistical characteristics (such as amplitude and root mean square value) of the observer residuals (output estimation error). However, this type of method has a fundamental flaw: changes in residual amplitude are the result of abrupt changes in the perturbation spectrum structure, rather than its essential characteristics. Sensor impulse interference or unmodeled constant offsets can also cause amplitude abrupt changes, leading to serious false triggering problems. More importantly, at the moment of switching, if the observer's internal state is not specially processed, a significant transient estimation error shock will occur due to the mismatch between the old and new dynamic models, severely disrupting the continuity of control. Existing technologies use state reset or rely on the asymptotic convergence of the observer itself to absorb this transient, which is engineeringly unacceptable in demanding high-speed control systems.

[0024] Therefore, the existing technical system implicitly contains a common technical problem: it assumes that the perception of disturbance dynamics must rely on the complete calculation and analysis of its energy spectrum, or at least on the macroscopic statistics of the residual signal (such as amplitude). This bias leads to a series of problems, including high computational load, insensitivity to structural changes, and susceptibility to unstructured disturbances. A deeper core problem lies in the fact that existing variable structure methods generally lack an effective mechanism to maintain the continuity of the estimated output at switching moments, inevitably introducing transient shocks. This fundamental abrupt change in the spectral structure of the disturbance poses a severe challenge to traditional ESOs designed based on linear time-invariant or fixed-order assumptions, causing their estimation performance to deteriorate sharply or even fail completely.

[0025] Figure 1 This is a flowchart illustrating an observer-based disturbance estimation method provided in an embodiment of this application. Figure 1 As shown, this observer-based perturbation estimation method may include the following steps: S110. Based on the corresponding state vector, determine the observation index values ​​corresponding to the high-order observer and the low-order observer respectively, and based on their respective target thresholds, determine whether there is a target observation index value among the observation index values, and obtain the corresponding judgment result; the target observation index value is the observation index value that exceeds the target threshold.

[0026] The state vector is a multi-dimensional data set describing the internal operating state of the observer. It includes the native measurable states of the controlled object, such as roll rate, and extended disturbance-related states, such as total disturbance and disturbance derivative. Within each control cycle, the observer updates the current cycle's state vector based on the previous cycle's state vector, system input, and measurement output through the state estimation equation, continuously outputting the latest state information. Specifically, the state vector of a low-order observer can be designed as follows: Where ω is the roll rate, d is the total perturbation state of the expansion, and T is the transpose. The state vector of the higher-order observer can be designed as follows: ,in These are the first and second derivatives of the perturbation, respectively.

[0027] The observation index values ​​are key indicators derived from the observer's state vector and used to quantitatively assess the health of the observer's operation. Specifically, they include two types of indicators: the residual magnitude of each observer and the trace of the estimation error covariance matrix of each observer. The residual magnitude is the absolute value of the deviation between the observer's output estimate and the system's measured value. The estimation error covariance matrix is ​​a matrix describing the statistical characteristics of the deviation between the state vector estimate and the true value. The trace, which is the sum of the elements on the main diagonal of the matrix, quantifies the overall magnitude of the estimation error and directly reflects the reliability of the state estimation. The smaller the trace, the more reliable the estimation result. In practical applications, the magnitude of the observation index values ​​directly reflects whether the observer is operating normally. If the residual magnitude is too large, it indicates a serious deviation between the observer's output estimate and the actual system state. If the trace of the covariance matrix is ​​too large, it indicates a deterioration in the statistical characteristics of the state estimation error, and the observer may be prone to divergence.

[0028] The target threshold refers to the critical value used to distinguish between normal operation and abnormal failure of an observer. Corresponding to the observed index values, it is also divided into two categories: residual amplitude alarm threshold and covariance trace threshold. The residual amplitude alarm threshold is for residual amplitude-type observed index values ​​and can be set to 3 to 5 times the root mean square (RMS) of the residuals during the stable flight phase, thus balancing the false alarm rate and the missed alarm rate based on the 3σ principle of statistical process control. The covariance trace threshold is for trace-type observed index values ​​of the estimated error covariance matrix and can be set to 3 to 5 times the trace of the covariance matrix during the stable flight phase, thus being consistent with the setting principle of the residual amplitude threshold and ensuring the reliability of the judgment. In practical applications, by comparing the two types of observed index values ​​of each observer with the corresponding target thresholds, if any index value exceeds the threshold, the observer is judged to be abnormal, that is, the target observed index value exists; if all index values ​​do not exceed the limit, the observer is judged to be normal.

[0029] Based on this, by determining the state vectors corresponding to the higher-order and lower-order observers within the current control cycle, the corresponding observation index values ​​are further determined. The observation index values ​​are then compared with the target threshold to determine whether a target observation index value exists, i.e., an observation index value exceeding the target threshold. If any index value exceeds the threshold, then a target observation index value exists, and the observer corresponding to the target observation index value is an observer with an anomaly. In practical applications, if an observer has an anomaly, the disturbance value estimated by that observer will also be inaccurate. The estimation error can be reduced and the accuracy of the disturbance estimation value improved by decreasing the weight corresponding to the abnormal observer.

[0030] S120. When the judgment result is that there is no target observation index value, determine the corresponding spectrum signal, and calculate the energy ratio based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, to obtain the high frequency energy ratio; substitute the high frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high frequency band are the signal frequency bands extracted from the spectrum signal.

[0031] Among them, the spectrum signal refers to the representation of the signal in the frequency domain, which can map the signal in the time domain to the frequency domain. In practical applications, the spectrum signal can be the frequency-amplitude distribution data obtained by the Fast Fourier Transform (FFT) after the residual signal of the observer is updated by sliding window and windowed. This intuitively reflects the frequency components and corresponding energy in the residual and directly presents the frequency composition in the residual signal, so as to determine the specific scenario of the disturbance spectrum being concentrated in a single peak or dispersed in multiple peaks. That is, the spectrum is distributed in a single peak under stable conditions and in a multi-peak distribution under extreme conditions.

[0032] The target frequency band and the high-frequency band are specific frequency ranges extracted from the spectrum signal for energy quantification. The target frequency band, also known as the band of interest, covers the entire frequency range where disturbances may exist and is the global range for energy calculation. For example, it is the 0-100Hz range set according to the attitude dynamics characteristics of an aircraft. The high-frequency band is the sub-range of the target frequency band that is higher than the estimated value of the main frequency peak, that is, the range of the target frequency band that is higher than the preset frequency peak value. It can be understood as the core range used to capture the high-frequency components of multi-peak disturbances. By extracting specific frequency bands, the global disturbance energy (target frequency band) and the high-frequency disturbance energy (high-frequency band) can be calculated separately, avoiding the interference of irrelevant frequencies (such as UHF noise) in the energy ratio calculation. Moreover, the energy ratio of the high-frequency band directly reflects the dispersion of the disturbance spectrum. The high-frequency band energy is extremely small under single-peak conditions and significantly increases under multi-peak conditions.

[0033] The energy ratio is the ratio of high-frequency band energy to the total energy of the target frequency band. It is a quantitative result obtained by "high-frequency energy ÷ total energy of the target frequency band", which transforms the energy distribution in the frequency domain into a single value and intuitively reflects the proportion of high-frequency components in the total disturbance.

[0034] The high-frequency energy ratio is the calculated result of the energy ratio. In practical applications, during the stable flight phase, the disturbance energy is highly concentrated in the main frequency peak, and the high-frequency energy ratio is close to 0. When the disturbance spectrum evolves into a multi-peak distribution, significant energy appears in the high-frequency region, and the high-frequency energy ratio will increase significantly. Therefore, this index can directly reflect the structural change of the disturbance spectrum from "concentrated" to "dispersed". It has a small computational load (only one FFT and piecewise integration are required), clear physical meaning, and is not sensitive to broadband interference such as sensor pulses, and has extremely high engineering practicality.

[0035] The target weight model is a pre-determined function model that can establish a continuous mapping relationship between the high-frequency energy ratio and the fusion weight. Specifically, it can be a smooth function with a value range of [0,1], such as the sigmoid hyperbolic tangent function. This enables the continuous output of the weight α of the low-order observer based on the real-time change of the high-frequency energy ratio, while the weight of the high-order observer is 1-α. This avoids the transient impact caused by hard switching of observers of different orders in existing technologies, takes into account the performance under all operating conditions, and ensures that when the high-frequency energy ratio is low (single peak), α≈1, at which time the low-noise advantage of the low-order observer is brought into play, while when the high-frequency energy ratio is high (multiple peak), α≈0, at which time the strong tracking advantage of the high-order observer is brought into play, thus achieving the optimal weight allocation under all operating conditions.

[0036] Dynamic low-order weights and dynamic high-order weights are the outputs of the target weight model. They are dynamic coefficients used to weight and fuse the perturbation estimates of low-order and high-order observers, satisfying "low-order weight + high-order weight = 1" to ensure the rationality of the fusion. Dynamic low-order weights correspond to the perturbation estimation weights of low-order observers, and dynamic high-order weights correspond to the perturbation estimation weights of high-order observers. The larger the dynamic weight corresponding to an observer, the more accurate the perturbation estimation of that observer is in the current scenario, that is, the more the observer is adapted to the real-time scenario. Thus, by increasing the dynamic weights, the importance of the perturbation estimation value corresponding to that observer is enhanced, so as to improve the accuracy of the final output target perturbation estimation value.

[0037] Based on this, when the judgment result is that there is no target observation index value, that is, the current observers are all operating normally and there is no error, the corresponding spectrum signal of the system at this time can be determined so as to extract the target frequency band based on the spectrum signal, and further determine the high frequency band in the target frequency band that is higher than the preset high frequency value. Then, based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, the energy ratio is calculated to obtain the high frequency energy ratio. The high frequency energy ratio is then substituted into the target weight model to obtain the dynamic low-order weights output by the model, so as to further determine the dynamic high-order weights.

[0038] S130. When the judgment result indicates that a target observation index value exists, determine the preset low-order weight and preset high-order weight according to the target observer category corresponding to the target observation index value.

[0039] The target observer category is a clearly defined type of abnormal observer based on the target observation index value exceeding the target threshold. It includes two categories: low-order observer category and high-order observer category. When the observation index value of a low-order observer exceeds the corresponding target threshold, the target observer category is determined to be a low-order observer, i.e., a low-order observer is abnormal. When the observation index value of a high-order observer exceeds the corresponding target threshold, the target observer category is determined to be a high-order observer, i.e., a high-order observer is abnormal.

[0040] Preset low-order weights and preset high-order weights are fixed weight coefficients predetermined for observer anomaly scenarios. They satisfy "preset low-order weight + preset high-order weight = 1", thereby achieving complete reliance on healthy observers and completely isolating abnormal observers. For example, in practical applications, if the target observer category is a low-order observer, then the preset low-order weight = 0 (or a very small safety value, such as 0.01), and the preset high-order weight = 1 (or a very large safety value, such as 0.99); while if the target observer category is a high-order observer, then the preset low-order weight = 1 (or a very large safety value, such as 0.99), and the preset high-order weight = 0 (or a very small safety value, such as 0.01).

[0041] Based on this, when the judgment result is that there is a target observation index value, that is, the current observer is not operating normally and there is an error, the specific observer with the error is determined according to the target observer category corresponding to the target observation index value, so as to determine the actual weight of each observer at this time according to the preset low-order weight and preset high-order weight.

[0042] S140. Based on the current corresponding low-order and high-order weights, the perturbation estimates output by the low-order and high-order observers are weighted and summed to obtain the target perturbation estimate.

[0043] Among them, low-order weights and high-order weights are coefficients used to allocate the contribution of low-order observers and high-order observers to the perturbation estimation values. The two are complementary, that is, the low-order weight is the weight corresponding to the low-order observer, and the high-order weight is the weight corresponding to the high-order observer. When the judgment result is that there is no target observation index value, the low-order weight is the dynamic low-order weight, and the high-order weight is the dynamic high-order weight; when the judgment result is that there is a target observation index value, the low-order weight is the preset low-order weight, and the high-order weight is the preset high-order weight. The weight directly determines the proportion of the perturbation estimation value of the corresponding observer in the final result. When the low-order weight is high, the advantages of the low-order observer of "low noise and high accuracy" are given priority; when the high-order weight is high, the advantages of the high-order observer of "strong tracking and wide bandwidth" are given priority.

[0044] The target disturbance estimate is the final quantized value of the total system disturbance obtained by weighted summation, which is then used for disturbance feedforward compensation. That is, by outputting the target disturbance estimate to the downstream control law and disturbance compensator, the compensator can generate a reverse control command based on the value to offset the damage to the system caused by the total disturbance such as model error and aerodynamic disturbance in advance.

[0045] Based on this, under stable operating conditions, this value has both low noise and high accuracy due to the dominance of low-order weights; under extreme disturbance conditions, this value can accurately track multi-frequency disturbances due to the dominance of high-order weights; when the observer is abnormal, this value relies entirely on the healthy observer to ensure that the control decision is not contaminated by distorted data.

[0046] Based on the feasible implementation of S110 described above, this application further provides a method for determining the observation index values ​​corresponding to the higher-order observer and the lower-order observer respectively based on the corresponding state vector, including: Based on the state vectors of the higher-order and lower-order observers in the current control cycle, the corresponding output estimates are determined respectively. Based on the roll rate measured by the gyroscope, the difference between the roll rate value and the output estimate is calculated to obtain the higher-order residual signal and the lower-order residual signal. Calculate the matrix traces of the estimation error covariance matrices corresponding to the higher-order and lower-order observers respectively to obtain the higher-order matrix traces and lower-order matrix traces; Determine the residual signal amplitude corresponding to each residual signal, and based on the corresponding residual signal amplitude and matrix trace, determine the observation index values ​​corresponding to the higher-order observer and the lower-order observer.

[0047] The output estimate is the real-time prediction of the core physical quantity of the system that can be directly measured, such as the roll rate, by the low-order and high-order observers based on their respective state vectors. Since the state vector contains ω as the roll rate, the predicted value of the roll rate of each observer can be determined, that is, the output estimate. This allows the output estimate of the observer to be directly compared with the roll rate value measured by the gyroscope, so as to obtain the residual signal corresponding to each observer.

[0048] The roll rate is a physical quantity that is directly measured by a gyroscope and describes the roll direction and speed of a highly dynamic moving body. It can be used as the reference true value for residual calculation.

[0049] The residual signal is the difference between the measured roll rate value of the gyroscope and the estimated value output by the observer, thus quantifying the observer's prediction bias. For two observers, the low-order residual signal reflects the deviation between the estimated value output by the low-order observer and the true value of the roll rate, i.e., the low-order residual signal; the high-order residual signal reflects the deviation between the estimated value output by the high-order observer and the true value of the roll rate, i.e., the high-order residual signal.

[0050] The estimation error covariance matrix is ​​a statistical matrix that describes the deviation between the estimated and actual values ​​of the observer's state. The trace of the matrix is ​​the sum of the elements on the main diagonal of the matrix. This allows for scalar quantification of the overall magnitude of the estimation error. The smaller the trace value, the more concentrated and stable the estimation errors of each state of the observer (roll rate, disturbance, etc.) are, and the higher the reliability of the state estimation. The larger the trace value, the more divergent the estimation error is, and the observer may be in an abnormal state.

[0051] Based on this, in practical applications, the health status of the observer is assessed by taking the absolute value of the residual signal, i.e., the residual signal amplitude. The larger the amplitude, the greater the prediction bias of the observer, and the more likely the operating status is abnormal. The trace value of the estimation error covariance matrix and the residual signal amplitude together constitute the observation index value, which is used to determine whether the observer is abnormal. When the trace value exceeds a preset threshold, it is determined that the estimation error of the corresponding observer is too large and needs to be adjusted by weight. In this way, since the residual signal is an instantaneous deviation and the covariance trace value is a statistical deviation, it can reflect the long-term estimation stability of the observer (e.g., if the residual is instantaneously normal but the trace value continues to increase, it indicates that the observer has a potential divergence risk). The combination of the two makes the health judgment more comprehensive.

[0052] Based on the feasible implementation of S110 described above, this application further provides a method for determining whether a target observation index value exists among the observed index values ​​based on their respective target thresholds, and obtaining the corresponding determination result, including: Based on the residual signal amplitude and matrix trace value corresponding to each observed index value, the corresponding comparison results are obtained by comparing each residual signal amplitude and matrix trace value with their respective target thresholds. If the comparison result shows that neither the residual signal amplitude nor the matrix trace value exceeds the corresponding target threshold, then the corresponding judgment result is determined to be that there is no target observation index value. If at least one of the residual signal amplitude and matrix trace exceeds the corresponding target threshold, the corresponding target observer category is determined, and the judgment result is determined to be that a target observation index value exists.

[0053] Based on this, the existence of target observation index values ​​can be determined by judging whether the residual signal amplitude and matrix trace value exceed the corresponding target threshold.

[0054] Based on the feasible implementation of S120 described above, this application further provides a method for determining the corresponding spectral signal, including: Based on any residual signal, the signal within the sliding window is updated to obtain the latest window signal, and a fast Fourier transform is performed on the latest window signal to obtain the spectrum signal.

[0055] The sliding window is a fixed-length time-domain data buffer used to store residual signals in real time within a continuous control cycle. It has a dynamic update mechanism of "new data enters the window and old data exits the window", that is, it receives only one new residual signal data point in each control cycle, and removes the oldest data point stored in the window at the same time, ensuring that a fixed number of continuous data points are always kept in the window.

[0056] The latest window signal is a set of continuous residual signals of fixed length stored in the window after the sliding window completes the update of new data entering the window and old data leaving the window. It is a complete data segment in the time domain. For example, when the window length is 256 data points, the latest window signal is a continuous sequence composed of "the residual signal of the current control cycle + the residual signal of the previous 255 control cycles", with the earliest old data removed.

[0057] Based on this, the sliding window converts discrete single-point signals into continuous time-domain segments by caching continuous data of a fixed length, providing the required input data for subsequent Fast Fourier Transform (FFT). The FFT requires continuous data segments to accurately analyze frequency components, while the dynamic sliding of the window ensures that the residual signal within the latest time period is analyzed each time, and can track the changes in the perturbation spectrum in real time. At the same time, the fixed length avoids the instability of spectrum analysis accuracy caused by fluctuations in the amount of data, ensuring the reliability of high-frequency energy ratio calculation.

[0058] Based on the feasible implementation of S120 described above, this application further provides a method that, before substituting the high-frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weights and dynamic high-order weights, includes: The hyperbolic tangent model is determined as the model framework corresponding to the target weight model, the initial weight model is obtained, and the corresponding perturbation signal is determined based on the simulation environment; Based on the disturbance signal, the disturbance estimate of the output of the initial weight model is determined, and the energy ratio parameter threshold and control parameters in the initial weight model are optimized according to the disturbance estimate to obtain the target weight model.

[0059] Among them, the hyperbolic tangent model is a functional framework built on the hyperbolic tangent function to establish a continuous mapping relationship between the high-frequency energy ratio and the fusion weight. It is the basic structure of the target weight model. The function curve is a smooth S-shape with no abrupt change points, which can realize the continuous transition of weight from 0 to 1. It fundamentally avoids the transient impact caused by the hard switching of observers of different orders for different practical scenarios in existing technologies.

[0060] Disturbance signals are artificially injected signals covering single-peak / multi-peak disturbance spectra in a simulation environment to simulate real flight conditions, in order to achieve disturbance simulation.

[0061] The energy ratio parameter threshold and control parameters are key adjustable parameters that need to be determined through optimization in the initial weight model, determining the mapping relationship between the weights and the high-frequency energy ratio. The energy ratio parameter threshold can be understood as the critical point for weight switching, which is the critical value of the high-frequency energy ratio. When the high-frequency energy ratio is consistent with the determined energy ratio parameter threshold, the weight α of the low-order observer is 0.5. At this time, the weights of the low-order observer and the high-order observer each account for 50%, which is the high-frequency energy ratio threshold that distinguishes between single-peak perturbations and multi-peak perturbations. The control parameter can be understood as the transition steepness parameter, which is used to control the rate at which the weight α of the low-order observer changes with the high-frequency energy ratio. The larger the value of the control parameter, the steeper the change of α near the energy ratio parameter threshold, that is, the more sensitive the weight switching. The smaller the value of the control parameter, the smoother the transition and the smoother the weight change.

[0062] Based on this, in order to achieve a continuous transition of weights from 0 to 1, the hyperbolic tangent function can be selected as the functional framework of the model. Furthermore, in the simulation environment, the energy ratio parameter threshold and control parameters in the initial weight model are optimized according to the disturbance estimate and actual disturbance value output by the initial weight model to obtain the target weight model.

[0063] Based on the feasible implementation of S120 described above, this application further provides a method for optimizing the energy ratio parameter threshold and control parameters in the initial weighting model according to the disturbance estimation value to obtain a target weighting model, including: Based on the simulation environment, the target disturbance value corresponding to the disturbance signal is determined, and the corresponding loss value is determined according to the loss function between the target disturbance value and the disturbance estimate. Based on the loss value and the preset loss threshold, the energy ratio parameter threshold and control parameters are optimized so that the optimized loss value is less than the loss threshold. The target weight model is determined based on the optimized energy ratio parameter threshold and control parameters.

[0064] The perturbation target value is the true quantized value corresponding to the artificially injected perturbation signal in the simulation environment, that is, the true value of the injected perturbation. It is the known benchmark data in the simulation environment, rather than the value estimated by the observer.

[0065] A loss function is a function used to quantify the magnitude of the deviation between the estimated disturbance value and the target disturbance value. For example, the integral squared error (ISE) can be selected as the loss function. The larger the difference between the estimated value and the true value, the larger the loss function value; the smaller the difference, the smaller the loss function value. The loss value is the single numerical result obtained by substituting the estimated disturbance value and the target disturbance value into the loss function.

[0066] The preset loss threshold is a pre-determined critical value for the loss, which is used to determine whether the weight model has been optimized and to avoid infinite loops in parameter optimization. When the optimized loss value is less than the preset loss threshold, it means that the model accuracy has reached the target and optimization stops; if it is always greater than the threshold, the parameters need to be optimized.

[0067] Based on this, by determining the loss function between the estimated disturbance value and the target disturbance value, the model parameters are adjusted, namely, the energy ratio parameter threshold and control parameters are optimized so that the optimized loss value is less than the loss threshold, thereby obtaining the optimized optimal energy ratio parameter threshold and control parameters, so as to obtain the target weight model.

[0068] Based on the feasible implementation of S130 described above, this application further provides a method for determining preset low-order weights and preset high-order weights according to the target observer category corresponding to the target observation index value, including: If the target observer category is a low-order observer, then the preset low-order weight is set to 0 and the preset high-order weight is set to 1. If the target observer category is a high-order observer, then the preset low-order weight is set to 1 and the preset high-order weight is set to 0.

[0069] Based on this, in practical applications, if the target observer category is a low-order observer, then the low-order weight is preset to 0 (or a very small safety value, such as 0.01), and the high-order weight is preset to 1 (or a very large safety value, such as 0.99); while if the target observer category is a high-order observer, then the low-order weight is preset to 1 (or a very large safety value, such as 0.99), and the high-order weight is preset to 0 (or a very small safety value, such as 0.01).

[0070] Please refer to Figure 2 , Figure 2 A schematic diagram illustrating the framework of an observer-based perturbation estimation method provided in this application embodiment; as shown Figure 2 As shown, in the data input and parallel observation layer, the sensor input module synchronously transmits the collected signals to the low-order ESO module and the high-order ESO module. The two modules run in parallel to generate their respective observation data. Subsequently, the core processing layer receives the output of the parallel observation layer, processes the data through the residual calculation module to obtain the residual signal, and then the spectral energy ratio calculation unit obtains the high-frequency energy ratio based on the residual signal. The dynamic weight fusion unit, combined with internal logic (including fault judgment rules), completes the weight fusion of the low-order and high-order observer outputs based on the high-frequency energy ratio. Finally, in the output and application layer, the fused result is passed to the disturbance estimation output unit to obtain the target disturbance estimation value, which is then passed to the control law disturbance compensator to ultimately serve external applications. The entire architecture realizes information interaction between modules through real-time data paths, achieving weight fusion and fault judgment.

[0071] Please refer to Figure 3 , Figure 3 This application provides a schematic diagram of an observer-based disturbance estimation method as an embodiment of the present application; Figure 3 As shown, the system covers two core stages: offline calibration and online real-time operation. It adopts a three-layer architecture. The data input and parallel observation layer includes a sensor input module and parallel low-order and high-order ESOs. The core processing layer includes a residual calculation module, a high-frequency energy ratio calculation unit, a dynamic weight fusion unit, and integrated health detection and weight adjustment logic. The output and application layer is a disturbance compensator that outputs the disturbance estimate to the control law. The offline calibration stage requires defining the dual-mode observer structure, determining the dynamic weight fusion function, and the health detection threshold. The online real-time operation stage sequentially executes the following steps: system initialization, parallel observer recursion and residual calculation, HFER calculation, dynamic weight fusion and fault management, and disturbance compensation output.

[0072] Please refer to Figure 4 , Figure 4 A schematic diagram of the calculation process of an observer-based disturbance estimation method provided in this application embodiment; as shown Figure 4As shown, the observation and acquisition and preprocessing stages involve data prediction, differencing, and denoising; full-band FFT analysis (capturing blade changes) is performed to complete frequency domain analysis and processing; frequency band energy is divided and calculated, and the energy of different frequency bands is planned and calculated; the high-frequency energy ratio (HFER) calculation step is entered to obtain the high-frequency energy ratio result; finally, the quantization is completed through the generation of dynamic fusion weights (including adaptive weights), clearly demonstrating the entire chain of steps from data preparation and frequency domain analysis to energy calculation and weight generation.

[0073] Please refer to Figure 5 , Figure 5 A logical schematic diagram of an observer-based disturbance estimation method provided in this application embodiment; as shown Figure 5 As shown, a system architecture comprising three modules—parallel health detection, fault detection and decision-making, and output—is illustrated. In the parallel health detection module, two observers, LD-ESO and HL-ESO, each receive input signals and synchronously complete state estimation, fault generation, and health assessment. The fault detection and decision-making module, through a fault detection and weight adjustment unit, combines the outputs of the two observers to perform fault judgment, weight calculation, and dynamic adjustment. The output module, through a fusion adjustment unit, uses a PID controller to achieve the final system fusion output.

[0074] Based on the above steps, it can be seen that this application, by introducing the physically clear and computationally lightweight characteristic quantity of "high-frequency energy ratio," achieves direct perception of the structural change in the disturbance spectrum from "energy concentration" to "energy dispersion." This enables the system to fundamentally distinguish between "spectral structure abrupt changes" requiring adaptive response and unstructured changes such as "sensor pulse interference" that should be suppressed. This solves the inherent problems of high false alarm rate and poor specificity in traditional monitoring methods based on residual amplitude or variance. Furthermore, the parallel observer + dynamic weight fusion architecture eliminates the instantaneous hard switching of observer order. By continuously adjusting the fusion weights, it ensures the continuity and smoothness of the final disturbance estimation output, eliminates switching transient impacts, avoids overshoot and phase jitter caused by traditional switching methods, and improves the disturbance compensation accuracy. The accuracy of the signal and the stability of the control system are ensured. In addition, the calculation of the high-frequency energy ratio only requires one FFT and a simple energy integral, and the computation of the weight function is also extremely small. Thus, it can operate stably within the millisecond control cycle, which fully meets the stringent real-time requirements of the embedded flight control computer of high-speed aircraft and has a solid engineering and practical foundation. The integrated fault detection and weight management mechanism deeply integrates fault handling into the core data fusion algorithm, giving the system an inherent intelligent fault tolerance capability. When any observer tends to diverge due to its own abnormality (such as sensor failure), the system can actively and intelligently adjust its contribution to the final output, realize front-end pollution isolation, effectively prevent fault propagation and system-level failure, and improve the survivability and safety reliability of the entire control system under abnormal conditions.

[0075] Figure 6 This is a schematic diagram of an observer-based disturbance estimation device provided in an embodiment of this application. Figure 6 As shown, this observer-based disturbance estimation device includes: a judgment module, a calculation module, a determination module, and a summation module; wherein: The judgment module is used to determine the observation index values ​​corresponding to the high-order observer and the low-order observer respectively based on the corresponding state vector, and to determine whether there is a target observation index value among the observation index values ​​according to their respective target thresholds, and to obtain the corresponding judgment result; the target observation index value is the observation index value that exceeds the target threshold. The calculation module is used to determine the corresponding spectrum signal when the judgment result is that there is no target observation index value, and to calculate the energy ratio based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, so as to obtain the high frequency energy ratio; and to substitute the high frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high frequency band are the signal frequency bands extracted from the spectrum signal; The determination module is used to determine the preset low-order weights and preset high-order weights based on the target observer category corresponding to the target observation index value when the judgment result indicates that a target observation index value exists. The summation module is used to perform a weighted summation of the perturbation estimates output by the low-order and high-order observers, respectively, based on the current corresponding low-order and high-order weights, to obtain the target perturbation estimate.

[0076] In this embodiment of the application, the determination module can also be specifically used for: Based on the state vectors of the higher-order and lower-order observers in the current control cycle, the corresponding output estimates are determined respectively. Based on the roll rate measured by the gyroscope, the difference between the roll rate value and the output estimate is calculated to obtain the higher-order residual signal and the lower-order residual signal. Calculate the matrix traces of the estimation error covariance matrices corresponding to the higher-order and lower-order observers respectively to obtain the higher-order matrix traces and lower-order matrix traces; Determine the residual signal amplitude corresponding to each residual signal, and based on the corresponding residual signal amplitude and matrix trace, determine the observation index values ​​corresponding to the higher-order observer and the lower-order observer.

[0077] In this embodiment of the application, the determination module can also be specifically used for: Based on the residual signal amplitude and matrix trace value corresponding to each observed index value, the corresponding comparison results are obtained by comparing each residual signal amplitude and matrix trace value with their respective target thresholds. If the comparison result shows that neither the residual signal amplitude nor the matrix trace value exceeds the corresponding target threshold, then the corresponding judgment result is determined to be that there is no target observation index value. If at least one of the residual signal amplitude and matrix trace exceeds the corresponding target threshold, the corresponding target observer category is determined, and the judgment result is determined to be that a target observation index value exists.

[0078] In this embodiment of the application, the calculation module can also be specifically used for: Based on any residual signal, the signal within the sliding window is updated to obtain the latest window signal, and a fast Fourier transform is performed on the latest window signal to obtain the spectrum signal.

[0079] In this embodiment of the application, other modules of this observer-based disturbance estimation device may also be specifically used for: The hyperbolic tangent model is determined as the model framework corresponding to the target weight model, the initial weight model is obtained, and the corresponding perturbation signal is determined based on the simulation environment; Based on the disturbance signal, the disturbance estimate of the output of the initial weight model is determined, and the energy ratio parameter threshold and control parameters in the initial weight model are optimized according to the disturbance estimate to obtain the target weight model.

[0080] In this embodiment of the application, other modules of this observer-based disturbance estimation device may also be specifically used for: Based on the simulation environment, the target disturbance value corresponding to the disturbance signal is determined, and the corresponding loss value is determined according to the loss function between the target disturbance value and the disturbance estimate. Based on the loss value and the preset loss threshold, the energy ratio parameter threshold and control parameters are optimized so that the optimized loss value is less than the loss threshold. The target weight model is determined based on the optimized energy ratio parameter threshold and control parameters.

[0081] In this embodiment of the application, the determining module can also be specifically used for: If the target observer category is a low-order observer, then the preset low-order weight is set to 0 and the preset high-order weight is set to 1. If the target observer category is a high-order observer, then the preset low-order weight is set to 1 and the preset high-order weight is set to 0.

[0082] Figure 7 This is a schematic diagram of a device for performing an observer-based disturbance estimation method according to an embodiment of this application. Figure 7 As shown, the device includes: The device may include one or more processors with processing cores, one or more computer-readable storage media such as memory, communication components, etc. The processor, memory, and communication components are connected via a bus.

[0083] In the specific implementation process, at least one processor executes computer execution instructions stored in memory, causing at least one processor to execute the above-described observer-based perturbation estimation method.

[0084] The specific implementation process of the processor can be found in the above method embodiments, and its implementation principle and technical effect are similar, so it will not be repeated here.

[0085] Furthermore, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0086] The memory may include Random Access Memory (RAM) and may also include Non-volatile Memory (NVM), such as at least one disk storage device.

[0087] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0088] In some embodiments, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps in any of the observer-based perturbation estimation methods described above.

[0089] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0090] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0091] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of program codes that can be loaded by a processor to execute steps in any of the observer-based disturbance estimation methods provided in embodiments of this application.

[0092] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0093] According to one aspect of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium.

[0094] Since the instructions stored in the storage medium can execute the steps in any of the observer-based disturbance estimation methods provided in the embodiments of this application, the beneficial effects that any of the observer-based disturbance estimation methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0095] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the appended claims.

[0096] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A disturbance estimation method based on an observer, characterized in that, The method includes: Based on the corresponding state vector, the observation index values ​​corresponding to the high-order observer and the low-order observer are determined respectively, and based on their respective target thresholds, it is determined whether there is a target observation index value among the observation index values, and the corresponding judgment result is obtained; the target observation index value is the observation index value that exceeds the target threshold. When the judgment result indicates that the target observation index value does not exist, the corresponding spectrum signal is determined, and the energy ratio is calculated based on the energy values ​​corresponding to the target frequency band and the high-frequency frequency band, respectively, to obtain the high-frequency energy ratio; the high-frequency energy ratio is substituted into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high-frequency frequency band are the signal frequency bands extracted from the spectrum signal; When the judgment result indicates that the target observation index value exists, a preset low-order weight and a preset high-order weight are determined according to the target observer category corresponding to the target observation index value. Based on the corresponding low-order and high-order weights, the perturbation estimates output by the low-order observer and the high-order observer are weighted and summed to obtain the target perturbation estimate.

2. The method according to claim 1, characterized in that, The step of determining the observation index values ​​corresponding to the higher-order observer and the lower-order observer based on the corresponding state vector includes: Based on the state vectors of the higher-order observer and the lower-order observer in the current control cycle, the corresponding output estimates are determined respectively, and based on the roll rate value measured by the gyroscope, the difference between the roll rate value and the output estimate is calculated to obtain the higher-order residual signal and the lower-order residual signal. Calculate the matrix traces of the estimation error covariance matrices corresponding to the higher-order observer and the lower-order observer respectively to obtain the higher-order matrix traces and the lower-order matrix traces; Determine the residual signal amplitude corresponding to each residual signal, and determine the observation index value corresponding to the higher-order observer and the lower-order observer respectively based on the corresponding residual signal amplitude and matrix trace value.

3. The method according to claim 2, characterized in that, The step of determining whether a target observation index value exists among the observed index values ​​based on their respective target thresholds, and obtaining the corresponding determination result, includes: Based on the residual signal amplitude and matrix trace value corresponding to each of the observed index values, compare each residual signal amplitude and matrix trace value with the target threshold corresponding to them to obtain the corresponding comparison result; If the comparison result is that neither the residual signal amplitude nor the matrix trace value exceeds the corresponding target threshold, then the corresponding judgment result is determined to be that the target observation index value does not exist. If at least one of the residual signal amplitude and the matrix trace exceeds the corresponding target threshold, then the corresponding target observer category is determined, and the judgment result is determined to be that the target observation index value exists.

4. The method according to claim 2, characterized in that, Determining the corresponding spectral signal includes: Based on any of the residual signals, the signal within the sliding window is updated to obtain the latest window signal, and the latest window signal is subjected to a fast Fourier transform to obtain the spectrum signal.

5. The method according to claim 1, characterized in that, Before substituting the high-frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weights and dynamic high-order weights, the method further includes: The hyperbolic tangent model is determined as the model framework corresponding to the target weight model, the initial weight model is obtained, and the corresponding perturbation signal is determined based on the simulation environment; Based on the disturbance signal, the disturbance estimate output by the initial weight model is determined, and the energy ratio parameter threshold and control parameters in the initial weight model are optimized according to the disturbance estimate to obtain the target weight model.

6. The method according to claim 5, characterized in that, The step of optimizing the energy ratio parameter threshold and control parameters in the initial weighting model based on the disturbance estimation value to obtain the target weighting model includes: Based on the simulation environment, the target disturbance value corresponding to the disturbance signal is determined, and the corresponding loss value is determined according to the loss function between the target disturbance value and the estimated disturbance value; Based on the loss value and the preset loss threshold, the energy ratio parameter threshold and the control parameter are optimized so that the optimized loss value is less than the loss threshold. The target weight model is determined based on the optimized energy ratio parameter threshold and the control parameters.

7. The method according to claim 1, characterized in that, The step of determining preset low-order weights and preset high-order weights based on the target observer category corresponding to the target observation index value includes: If the target observer category is the low-order observer, then the preset low-order weight is determined to be 0 and the preset high-order weight is determined to be 1. If the target observer category is the higher-order observer, then the preset low-order weight is determined to be 1 and the preset high-order weight is determined to be 0.

8. An observer-based disturbance estimation device, characterized in that, The device includes: The judgment module is used to determine the observation index values ​​corresponding to the high-order observer and the low-order observer respectively based on the corresponding state vector, and to determine whether there is a target observation index value among the observation index values ​​according to their respective target thresholds, so as to obtain the corresponding judgment result; the target observation index value is the observation index value that exceeds the target threshold. The calculation module is used to determine the corresponding spectrum signal when the judgment result is that the target observation index value does not exist, and to calculate the energy ratio based on the energy values ​​corresponding to the target frequency band and the high frequency band respectively, so as to obtain the high frequency energy ratio; and to substitute the high frequency energy ratio into the target weight model to obtain the corresponding dynamic low-order weight and dynamic high-order weight; the target frequency band and the high frequency band are the signal frequency bands extracted from the spectrum signal; The determination module is used to determine a preset low-order weight and a preset high-order weight according to the target observer category corresponding to the target observation index value when the determination result is that the target observation index value exists; The summation module is used to perform a weighted summation of the perturbation estimates output by the low-order observer and the high-order observer according to the corresponding low-order weight and high-order weight, so as to obtain the target perturbation estimate.

9. A computer device, characterized in that, include: One or more processors; Memory; One or more programs, wherein the one or more programs are stored in memory and configured to be executed by one or more processors, the one or more programs being configured to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that can be called by a processor to perform the method as described in any one of claims 1 to 7.

Images