Radar lifting control method and system based on meteorological monitoring

Abstract

The invention discloses a radar lifting control method and system based on meteorological monitoring, and relates to the technical field of radar lifting control, and the method comprises the steps: completing the switching of a power supply and communication after a radar is powered on, initializing a controller, collecting the data of a meteorological station, and generating a future fusion wind speed in real time through a Kalman filtering physical model and a residual neural network; future fused wind speed is converted into wind pressure for evaluation, the risk degree is judged according to the evaluation result, early warning is given out, and the controller is preheated to enter a lifting preparation state. The input stability is improved through meteorological data sliding window smoothing and feature extraction, wind speed dynamic prediction and uncertainty quantification are achieved through XGBoost prediction and residual variance estimation, the time sequence consistency and robustness are enhanced through remote API interpolation correction and adaptive extended Kalman filtering, residual correction is conducted through a neural network, the prediction precision is improved, and the prediction accuracy is improved. And a reliable decision basis is provided for radar lifting control.

Description

Technical Field

[0001] This invention relates to the field of radar elevation control technology, and in particular to a radar elevation control method and system based on meteorological monitoring. Background Technology

[0002] With the widespread application of radar equipment in mine slope monitoring, the safety and stability of its operating environment have received increasing attention. Traditional radar lifting mechanisms are mostly controlled based on fixed rules or human experience, such as determining whether to raise or lower the equipment according to a preset wind speed threshold. While these methods are simple to implement, they are insufficient in responding to complex and changing atmospheric environments, making it difficult to guarantee the safety of the equipment under extreme weather conditions. In recent years, with the development of sensor technology, numerical weather prediction models, and intelligent control theory, the role of meteorological monitoring and forecasting in the safe operation of equipment has become increasingly prominent. Collecting real-time meteorological elements from split-type meteorological stations and combining them with machine learning models and filtering methods to predict short-term wind speed and direction has become an important trend. At the same time, advanced control methods such as Model Predictive Control (MPC) and Quadratic Programming (QP) optimization techniques are gradually being applied to the dynamic scheduling and constraint solving of complex systems, enabling equipment to achieve better dynamic responses under uncertain disturbance environments. Overall, related research and engineering practices are evolving from single threshold control to a comprehensive control approach based on meteorological monitoring, predictive modeling, and intelligent optimization decision-making.

[0003] However, existing technologies still have significant shortcomings. First, current radar rise and fall control systems do not fully utilize meteorological information, often relying on single meteorological elements or rough averages, failing to effectively integrate local observation data with long-distance weather forecast data, resulting in insufficient prediction accuracy. Second, in wind speed prediction, traditional methods are usually based on linear regression or empirical formulas, which struggle to capture the nonlinear dynamic characteristics and random disturbances of the wind field. Even with advanced models, their characterization of prediction uncertainties is often insufficient, failing to provide a reliable basis for subsequent risk assessment and control optimization. Therefore, radar rise and fall technology for meteorological monitoring has shortcomings in data fusion and prediction accuracy. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a radar rise and fall control method and system based on meteorological monitoring, which solves the problems of data fusion, prediction accuracy and intelligent control.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a radar elevation control method based on meteorological monitoring, comprising, After the radar is powered on, it completes the power and communication switching, initializes the controller, collects weather station data, and generates the future fused wind speed in real time through the Kalman filter physical model and residual neural network. The future wind speed is converted into wind pressure for assessment. Based on the assessment results, the risk level is determined and an early warning is issued. The controller is then preheated and put into the lifting preparation state. Based on the warning, the controller is activated to perform a descent operation and lock the equipment to prevent it from rising. When the environment meets the safety conditions again, the lock is released and the equipment is raised through the controller.

[0007] As a preferred embodiment of the radar lift control method based on meteorological monitoring described in this invention, the following steps are taken: Meteorological station data is collected, and a future fused wind speed index is generated in real time using a Kalman filter physical model and a residual neural network. On-site meteorological data is collected from a split-type meteorological station at a fixed sampling frequency. The operating variables of the radar lift equipment are collected through internal sensors. The instantaneous wind speed values ​​of the most recent N sampling times are stored in a circular buffer using a sliding window. The average wind speed within the time window at each time point is calculated to construct an average wind speed sequence. The first-order gradient feature is obtained by calculating the difference between instantaneous wind speeds at adjacent times. The input feature vector is constructed by combining meteorological data and operational variables as external influencing factors. The data is input into an XGBoost regression model for prediction. The mean wind speed prediction is calculated for each prediction step, and then compared with actual wind speed observations. Compared with the average wind speed forecast Subtraction calculation of predicted residuals A residual circular buffer is established, and the variance of the predicted residual is used. As an estimate of the prediction uncertainty, the final output prediction pair ; Retrieve the mean wind speed forecast at discrete times from a remote API. The prediction variance is calculated by subtracting the historical API wind speed forecast mean from the actual measured value. Furthermore, piecewise cubic Hermite interpolation with monotonically preserved parameters is used to transform discrete-time meteorological forecast data into second-by-second sequences, and the prediction variance is analyzed. Perform dilation correction to obtain the interpolated corrected remote API variance. Assuming the remote API data is available, output the interpolated remote API prediction. Furthermore, an adaptive extended Kalman filter method is used to recursively estimate the wind speed state, defining a state vector. The state equations are in the form of nonlinear damping and process noise is incorporated, based on the state vector. The observation equation is defined, and the filtering recursion follows a two-stage process of "prediction-update". In the prediction stage, the prior state estimate vector at the current time is calculated through a nonlinear state transition function. The prior error covariance matrix is ​​obtained by propagating the covariance based on the Jacobian matrix. Based on the prior error covariance matrix and observed values Calculate the Kalman gain to obtain the Kalman gain matrix. The posterior state estimate vector is obtained by correcting the prior state estimate vector and the prior error covariance matrix at the current time. and posterior error covariance matrix , as the initial condition for prediction at the next moment; In the EKF update step, the measured wind speed at time k is used. With predicted wind speed The observation residuals are obtained by subtraction. Perform measurement noise covariance update based on observed residuals. and the updated measurement noise covariance Correct the noise covariance during the process and construct the input vector of the neural network. , input vector Input to neural network Output state correction and the posterior state estimation vector The final state estimate is obtained by summing the residuals. Estimate based on the final state As the initial state, the nonlinear state transition function Continuous stacking Step by step, the predicted state at each step is iterated by substituting the predicted state of the previous step into the nonlinear state transition function to obtain the future state. Step state prediction vector Simultaneously, the prediction error covariance matrix is ​​recursively derived. Combined with future state prediction vector Output predicted wind speed and variance. The XGBoost regression model prediction results are weighted with the remote API prediction results to obtain the fusion model's wind speed prediction value and variance. When the remote API is unavailable, it degenerates into a single model, and the model forecast results are displayed. The predicted wind speed is then fused with the prediction results from the adaptive extended Kalman filter method using a second weighted fusion method. variance is .

[0008] As a preferred embodiment of the radar rise and fall control method based on meteorological monitoring described in this invention, the step of converting future fused wind speed into wind pressure for evaluation refers to, since the actual wind speed is greater than or equal to 0, correcting the wind speed distribution parameters based on a left-truncated normal distribution, calculating the standardized cutoff point Z, and setting a threshold. When the standardized cutoff point Z is less than the threshold If the truncation effect is negligible, no correction is needed; otherwise, parameter correction is performed. For cases requiring correction, the standard normal density function is calculated. and cumulative distribution function Based on the existing derivation formula for truncated normal moments, the mean and variance of wind speed are corrected, and the one-sided confidence quantile at a given significance level 0 is calculated using the inverse function of the monotonic approximation function. The mean predicted wind pressure for each prediction step is calculated based on the corrected mean and variance of wind speed. Based on the corrected variance The Delta method is used to propagate the uncertainty in wind speed prediction to wind pressure, thus obtaining the variance in wind pressure prediction. Variance prediction by wind pressure Define wind pressure standard deviation By one-sided confidence quantiles Calculate the one-sided confidence maximum value of wind pressure at a significance level of o. Set the critical wind pressure threshold The maximum value of wind pressure at the significance level o Greater than or equal to the critical wind pressure threshold When this happens, it is determined to enter a "high-risk" state, and debouncing logic is used, that is, within the range If two or more prediction points exceed the threshold consecutively, the risk level is immediately confirmed as "high risk"; otherwise, wait for the next prediction cycle for verification.

[0009] As a preferred embodiment of the radar elevation control method based on meteorological monitoring described in this invention, the step of issuing an early warning based on the assessment result and preheating the controller to enter the elevation preparation state refers to the detection of the predicted step size. When the risk level is high, the on-site audible and visual warning device should be activated immediately to provide sound warnings, optical warnings, and human-machine interface prompts. The warning information should be packaged into a standardized data format and sent to the remote monitoring platform. At the same time as issuing the warning, the MPC control should be put into the ready state.

[0010] As a preferred embodiment of the radar elevation control method based on meteorological monitoring described in this invention, wherein: the start controller executes the descent operation and locks the equipment to prevent elevation by collecting and confirming the physical parameters of the equipment and its mounting components; at each control moment, the state vector is defined as follows: And by using the predicted wind pressure average Multiply by the equivalent windward area O to obtain the equivalent wind load. Equivalent wind load Dividing by the mass m to obtain the linear acceleration a, the equivalent wind load is then converted. Multiply by the torque arm g, then divide by the moment of inertia Y to obtain the angular acceleration. Based on linear acceleration a and angular acceleration Combined with the construction of external perturbation vector and define control inputs To achieve the desired descent rate, Model Predictive Control (MPC) employs a discrete state-space model to predict the future state vector evolution. MPC solves this problem using a finite-time optimization approach, with an objective function S. The MPC optimization must satisfy input constraints, tilt angle constraints, and soft wind pressure constraints. MPC solves a first-order quadratic programming problem to optimize the objective function S. Upon successful solution, MPC outputs the optimal control input command per second. During the descent process, the equipment descends until it reaches the minimum height or receives a manual stop command. Once the descent stops, the red "Do Not Lift" indicator light illuminates, triggering both mechanical and software locking to prevent the equipment from lifting immediately. The locking status is then reported to the remote platform, and the completion time is recorded.

[0011] As a preferred embodiment of the radar lifting control method based on meteorological monitoring described in this invention, when the environment meets the safety conditions again, the lock is released and the prediction of the future period is evaluated by the controller's lifting command. If the meteorological conditions, control command conditions, and wind pressure prediction conditions are met simultaneously, the system switches to "lifting mode". At this time, the control input is the lifting speed. During the lifting process, the wind pressure prediction conditions are continuously evaluated until the equipment is fully lifted into position.

[0012] As a preferred embodiment of the radar rise and fall control method based on meteorological monitoring described in this invention, the radar completes the power and communication switching after power-on and initializes the controller. After the radar system is powered on, the power status is detected and a communication link is established. After the power and communication are stable, the core hardware is started and the required models and parameters are loaded and run. During the loading process, the file version, hash value and size are checked, and a remote meteorological API channel is established to receive weather forecast data. When the remote meteorological API channel is unavailable, the local micro-area wind field extrapolation module is activated and the degradation status flag is recorded.

[0013] Secondly, the present invention provides a radar elevation control system based on meteorological monitoring, comprising, The power-on startup module is used to complete the power communication switch when the radar is powered on, load and verify the prediction model and control parameters, and establish a remote meteorological channel. The weather forecasting module is used to collect multi-element data in real time from the weather station, calculate the average wind speed using a sliding window, and combine XGBoost and EKF for fusion forecasting. The wind pressure assessment module is used to correct the mean and variance of wind speed based on a truncated normal distribution, derive the mean and variance of predicted wind pressure, and assess extreme risks. The risk warning module is used to trigger an alarm on-site by means of on-site sound and light devices when the wind pressure prediction exceeds the threshold to trigger a high risk, and to push the MPC control into the preparation state. The control and locking module is used for MPC to construct QP optimization solution for descent control input. After the device executes descent, it triggers double locking and records and reports the status.

[0014] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the radar rise and fall control method based on meteorological monitoring as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the radar rise and fall control method based on meteorological monitoring as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By integrating XGBoost regression, residual cyclic buffering, extended Kalman filtering, and meteorological API interpolation correction multi-source fusion methods, this invention achieves high-precision prediction and uncertainty estimation of future wind speed. Based on the wind speed-wind pressure mapping and extreme risk assessment mechanism with truncated normal distribution, it can more scientifically quantify the risk level. Finally, by combining the MPC and QP solution framework, the risk assessment results are directly converted into real-time control commands, effectively improving the safety and robustness of the radar rise and fall system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a radar elevation control method based on meteorological monitoring in Example 1.

[0019] Figure 2 This is a schematic diagram of a radar lift control system based on meteorological monitoring in Example 1.

[0020] Figure 3This is a flowchart of the MPC control and lifting execution process in Example 1. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1, referring to Figures 1-3 This is the first embodiment of the present invention, which provides a radar elevation control method based on meteorological monitoring, including the following steps: S1. After the radar is powered on, it completes the power and communication switching, initializes the controller, collects weather station data, and generates the future fused wind speed in real time through the Kalman filter physical model and residual neural network. Specifically, after the radar is powered on, it completes the power and communication switching and initializes the controller. After the radar system is powered on, it detects the power status. When an abnormal power supply is detected, it automatically switches to the backup generator for power supply, records alarm information, and establishes a communication link (4G or fiber optic). After the power and communication are stable, the core hardware (main controller and industrial PC) is started, and the required models and parameters are loaded and run, including XGBoost prediction model files, Kalman filter parameters, Model Predictive Control (MPC) coefficients (MPC simulates the future to select the optimal and safe action to execute), and control constraint parameters. During the loading process, the file version, hash value, and size are checked, and a remote meteorological API channel (such as NOAA, ECMWF, Windy, etc., which directly provides wind speed prediction values ​​for future moments) is established to receive weather forecast data. When the remote meteorological API channel is unavailable, the local micro-area wind field extrapolation module (based on the Lagrange method) is activated, and the degradation status flag is recorded.

[0025] By automatically switching and recording alarms in case of power failure, the system ensures continuous power supply and fault traceability. By establishing a stable communication link, remote monitoring and emergency command issuance are achieved. After the power supply and communication stabilize, the main control hardware is quickly started and the model and parameters are loaded. The integrity of the files is verified to ensure the reliability of prediction and control. At the same time, the remote meteorological API is connected to provide macro forecasts. When the API is unavailable, the forecasts degenerate to local Lagrange extrapolation and the degradation status is recorded to prompt the system to enter conservative control. This achieves multiple redundancies and protections from the hardware layer to the data layer, ultimately providing safe, continuous and reliable operation guarantee for radar rise and fall control.

[0026] Furthermore, meteorological station data is collected, and a future fused wind speed index is generated in real time using a Kalman filter physical model and residual neural network. Real-time meteorological data (instantaneous wind speed, rainfall, temperature, relative humidity) is collected from the split-type meteorological station at a fixed sampling frequency. Operating variables of the radar lifting equipment (blade pitch angle, power generation, rotational speed, etc.) are collected through internal sensors (angle encoder, current / voltage sensor, speed sensor, etc.). To ensure data continuity and real-time performance, the instantaneous wind speed values ​​of the most recent N sampling times are stored in a circular buffer using a sliding window, and the average wind speed within the time window is calculated for each moment.

[0027] in, Let be the average wind speed at time k, and i be the sampling time of the time series. It is the instantaneous wind speed value collected at sampling time i; An average wind speed sequence is constructed based on the average wind speed within the time window at each moment, and the first-order gradient feature is obtained by calculating the difference between instantaneous wind speeds at adjacent moments. Further dynamic change characteristics are extracted to reflect the changing trend of wind speed. Cumulative rainfall is obtained by summing up rainfall data obtained from the split-type weather stations.

[0028] in, It is the cumulative rainfall at time k. The instantaneous rainfall value collected at sampling time i is N, and N is the total number of sampling times. The temperature difference is calculated by subtracting the collected temperature data from the historical database at the same time 24 hours prior. The dew point temperature (the temperature at which air, with constant water vapor content and air pressure, is cooled to saturation (i.e., relative humidity reaching 100%)) is calculated based on the collected relative humidity data and temperature data.

[0029]

[0030] in, It is the dew point temperature at time k. It is the relative humidity at time k. It is the temperature at time k. This involves converting relative humidity from a percentage to a decimal form so that it can be used in logarithmic calculations. It is the natural logarithm function. , These are the empirical constants for temperature and humidity, respectively, derived from the Magnus formula. Setting thresholds via grid search The relative humidity at time k Greater than the threshold If the humidity is high at a given moment, then record that moment as a high humidity period. Simultaneously, record the duration of the high humidity and apply the first-order gradient feature. Combining meteorological data (cumulative rainfall, temperature difference) The input feature vector is constructed using the dew point temperature, high humidity, and duration of high humidity, along with the operating variables of the radar lifting equipment, as external influencing factors. The data is input into the XGBoost regression model for prediction, and the mean wind speed prediction is calculated for each prediction step.

[0031] in, It is the future moment The average wind speed forecast, The XGBoost regression model is in the prediction step size The mapping function below; Based on actual wind speed observations Compared with the average wind speed forecast Subtraction calculation of predicted residuals A residual circular buffer of fixed length M is established for each prediction step, and the variance of the prediction residuals is used as an estimate of the prediction uncertainty.

[0032] in, In predicting step size The predicted variance is below. Here, j is the residual mean, and j is the index number of the residual sample in the residual circular buffer. It is the prediction start time when the j-th residual sample is generated. It is at the predicted start time The prediction step size made The predicted residuals; residual mean The calculation formula is:

[0033] Finally, at time k, output for future times. The prediction To enhance prediction stability, the mean wind speed prediction at discrete times is obtained from a remote API. The prediction variance is calculated by subtracting the historical API wind speed forecast mean from the actual measured value. The discrete-time weather forecast data is transformed into a second-by-second sequence using monotonically preserved piecewise cubic Hermite interpolation (PCHIP). Considering the additional uncertainty in the interpolation process, the prediction variance is... After performing dilation correction, the formula for calculating the variance after interpolation correction is:

[0034] in, β is the interpolated and corrected remote API variance, and β is the interpolation uncertainty coefficient, which is obtained by calibration using historical weather forecasts and measured residuals. Assuming the remote API data is available, output the interpolated remote API prediction. Furthermore, the adaptive extended Kalman filter (NCA-EKF) method is used to recursively estimate the wind speed state, defining a state vector. Instantaneous wind speed Combined with wind speed change rate (Calculated using the finite difference method based on wind speed data from two adjacent moments), the state equation is in the form of nonlinear damping, and process noise is incorporated:

[0035] in, It is the prediction of the state at time k+1 based on the information at time k. It is a nonlinear state transition function, where T is the sampling period, α is the empirical decay coefficient (set experimentally), and sing(·) is the sign function. It's process noise. It is the process noise covariance; Based on state vector Define the observation equation:

[0036] in, Here, H represents the observed values, and H is the observation matrix. It measures noise. It measures the noise covariance; The filtering recursion follows a two-stage process of "prediction-update". In the prediction stage, the prior state estimate vector at the current time is calculated through a nonlinear state transition function. And propagate the covariance based on the Jacobian matrix:

[0037] in, It is the prior error covariance matrix at time k. It is the posterior error covariance matrix at time k-1. It is a transpose operation. It is the noise covariance of the process at time k-1. It is the Jacobian matrix at time k-1; Jacobian matrix The calculation formula is:

[0038] in, It is a nonlinear state transition function. It is the sign of a partial derivative; Based on the prior error covariance matrix and observed values Calculate the Kalman gain:

[0039] in, H is the Kalman gain matrix, and H is the observation matrix. It is a transpose operation. It measures the noise covariance; Based on Kalman gain matrix Correct the prior state estimate vector at the current time step. and prior error covariance matrix :

[0040]

[0041] in, It is the posterior state estimate vector at time k. These are predicted observations, among which, Let J be the posterior error covariance matrix at time k, and J be the identity matrix. Obtain the posterior state estimation vector pair As the initial condition for the prediction of the next moment, in the EKF update step, the measured wind speed at time k is used. With predicted wind speed The observation residuals are obtained by subtraction. Perform measurement noise covariance update:

[0042] in, This is the measurement noise update value at time k. It is the noise update value of the previous moment, and b is the forgetting factor; Based on the observed residuals and the updated measurement noise covariance Correction process noise covariance:

[0043] in, It is the process noise estimate at time k. This is the estimated process noise value from the previous time step. It is a smoothing factor; After completing the EKF update step, construct the input vector of the neural network. (Posterior state estimation vector) Kalman gain matrix Observation residuals (Wind direction angle, cumulative rainfall, temperature, relative humidity), input vector Input to neural network Output state correction and the posterior state estimation vector The final state estimate is obtained by summing the residuals. Estimate based on the final state As the initial state, the nonlinear state transition function Continuous stacking Step by step, the predicted state at each step is iterated by substituting the predicted state of the previous step into the nonlinear state transition function to obtain the future state. Step state prediction vector Simultaneously, the error covariance matrix is ​​recursively derived:

[0044] in, It is the prediction error covariance matrix. It is the Jacobian matrix at the prediction point. It is the process noise covariance; Based on future state prediction vector and prediction error covariance matrix Output predicted wind speed and variance:

[0045] in, It is a moment The average predicted wind speed, H is the variance of the predicted wind speed, i.e., the uncertainty of the wind speed prediction, and H is the observation matrix. The wind speed predictions and variance of the fusion model are obtained by weighting the XGBoost regression model predictions with the remote API predictions:

[0046] in, In predicting step size The predicted variance is below. It is the interpolated and corrected variance of the remote API. In predicting step size The average wind speed forecast, Is it a remote API at any time? Interpolation prediction; When the remote API is unavailable, it degenerates into a single model:

[0047] Model prediction results The predicted wind speed is then fused with the prediction results from the Adaptive Extended Kalman Filter (NCA-EKF) method using a second weighted fusion method. variance is .

[0048] Real-time acquisition and smoothing of meteorological data are achieved through fixed sampling and sliding windows, ensuring input continuity and noise resistance. Wind speed gradients are extracted differentially and features are constructed based on operational conditions. Short-term wind speed forecasting is achieved using XGBoost, improving responsiveness to sudden dynamic changes. Prediction variance is estimated using a residual circular buffer, quantifying uncertainty at the step size resolution. Furthermore, PCHIP interpolation and variance inflation correction are employed, combined with remote meteorological API data, to ensure temporal consistency and confidence reliability of the forecast. At the forecasting level, an adaptive extended Kalman filter is used to recursively deduce the state, utilizing… The system dynamically adjusts measurement and process noise based on observation residuals and combines this with neural network compensation for structural errors to obtain a more stable and accurate state estimate. Through state iteration and compounding, multi-step predictions are generated to ensure consistent propagation of the prediction mean and variance over time. Finally, by weighted fusion of XGBoost predictions, API correction results, and EKF outputs, a wind speed prediction mechanism with both accuracy and robustness is formed. When external forecasts are missing, the system automatically degenerates into local predictions to ensure continuity. This provides high-confidence wind speed and wind pressure predictions for radar rise and fall, reduces false alarm and missed alarm rates, and enables early prediction of extreme weather events.

[0049] S2. The future wind speed is converted into wind pressure for assessment. Based on the assessment results, the risk level is determined and an early warning is issued. The controller is preheated and put into the lifting preparation state. Specifically, converting future wind speed into wind pressure for evaluation means that since the actual wind speed is greater than or equal to 0, a normal distribution is directly adopted. This will introduce deviations in low-wind-speed regions, requiring truncation correction. The wind speed distribution parameters are corrected based on a left-truncation normal distribution, and the standardized cutoff point is calculated.

[0050] Where Z is the standardized cutoff point, It is the standard deviation of the fused predicted wind speed, which is the square root of the variance of the final fused predicted wind speed. Setting thresholds through Bayesian optimization When the standardized cutoff point Z is less than the threshold If the truncation effect is negligible, no correction is needed; otherwise, parameter correction is performed. For cases requiring correction, the standard normal density function and cumulative distribution function are calculated.

[0051] in, It is the standard normal probability density function. It is the cumulative distribution function. It is a mathematical constant, and e is the base of the natural logarithm. These are the normalization coefficients of the standard normal distribution, where t is the integral variable. It is negative infinity; Based on the existing derivation formula for the truncated normal distribution moments, the mean and variance of wind speed are corrected:

[0052] in, , These are the corrected mean and variance of wind speed, respectively. In extreme risk assessment, the one-sided confidence quantile at a given significance level o is calculated using the inverse function of the monotonic approximation function:

[0053] in, It is a monotonic approximation function, where q is a standardized variable. These are parameters, determined by the optimization algorithm. It is a one-sided confidence quantile. It is the inverse function of the approximate function; Calculate the mean predicted wind pressure for each prediction step based on the corrected mean and variance of wind speed:

[0054] in, It is the predicted average wind pressure. It is air density. It is the wind pressure coefficient; Based on the corrected variance The uncertainty in wind speed prediction is calculated by using the Delta method (variance propagation approximation) to propagate this uncertainty to wind pressure, resulting in the wind pressure prediction variance, which is the wind pressure prediction uncertainty.

[0055] in, It is the variance of wind pressure prediction; Variance prediction based on wind pressure Define wind pressure standard deviation The square root of the wind pressure forecast variance is used to assess extreme risks, through one-sided confidence quantiles. Calculate the one-sided confidence maximum value of wind pressure at a significance level of o:

[0056] in, It is the maximum value of the predicted wind pressure at a significance level of o, where o is the significance level; Setting critical wind pressure thresholds using fuzzy logic The maximum value of wind pressure at the significance level o Greater than or equal to the critical wind pressure threshold If the condition is detected at a certain time, it is determined to enter a "high-risk" state. To avoid false alarms caused by transient anomalies, debouncing logic is used, that is, within the interval... If two or more prediction points exceed the threshold consecutively, the risk level is immediately confirmed as "high risk". Otherwise, the prediction cycle will be repeated for verification to reduce the false alarm rate.

[0057] By applying a left-truncation normal correction to the wind speed prediction distribution, consistent handling of physical constraints is achieved, avoiding deviations caused by negative probabilities in low-wind-speed areas. The mean and variance are rigorously corrected using the truncation normal moment formula, ensuring the reproducibility and reliability of statistical results. A monotone approximation function is used to inversely calculate the one-sided confidence quantiles, enabling stable and rapid calculation of tail quantiles in extreme risk assessment. Combined with the Delta method, the uncertainty of wind speed prediction is propagated to the wind pressure level, obtaining the wind pressure mean and variance with confidence intervals, providing a basis for actual risk indicators. By calculating the maximum confidence value of wind pressure at the significance level, a conservative load limit under extreme conditions is constructed. Combined with an anti-shake mechanism, risk assessment can smoothly transition in the critical zone while suppressing false alarms caused by single-point noise.

[0058] Furthermore, based on the assessment results, a risk level is determined, an early warning is issued, and the preheating controller enters the lifting preparation state when the predicted step size is detected. When the risk level is high, the on-site audible and visual warning device is immediately activated to provide audible warnings (the buzzer emits intermittent sounds at an intensity of 85 dB), optical warnings (the red warning light flashes at a frequency of 2 Hz), and human-machine interface (HMI) display prompts. The warning information (equipment ID, first time exceeding the limit, current fused predicted wind speed, corresponding confidence upper limit wind pressure, data acquisition timestamp) is packaged into a standardized data format and sent to the remote monitoring platform. At the same time as issuing the warning, the MPC control is pushed into the preparation state, including setting flag bits, preloading model parameters, and preprocessing QP factor decomposition.

[0059] When the predicted step size enters a high-risk phase, an audible, visual, and HMI multi-channel early warning is immediately triggered to ensure that personnel can perceive the risk in a timely manner. At the same time, the early warning information is standardized and uploaded to a remote platform to achieve remote monitoring and traceability. MPC control is pre-loaded and QP is decomposed to shorten the execution delay. The whole system forms a closed loop of risk identification, personnel reminder, remote sharing, and control pre-setting, which significantly improves the safety and response efficiency of radar lift control.

[0060] S3. Based on the warning, the controller is activated to perform a descent operation and the equipment is locked to prevent it from rising. When the environment meets the safety conditions again, the lock is released and the device is raised through the controller.

[0061] Specifically, the controller is activated to perform a descent operation and lock the equipment to prevent it from rising. The physical parameters of the equipment and its mounting components are collected and confirmed (mass, torque arm, and moment of inertia are obtained through manufacturer data or through on-site static weighing, torsion / dynamic testing, and system identification). At each control moment, the state vector is defined as follows: ,in, It is a state vector. It refers to the height position of the equipment. It is vertical velocity. It is the tilt angle of the equipment. It is the tilt angle and angular velocity. It is a transpose operation, and it is performed by taking the average predicted wind pressure. Multiply by the equivalent windward area O to obtain the equivalent wind load. Equivalent wind load Dividing by the mass m to obtain the linear acceleration a, the equivalent wind load is then converted. Multiply by the torque arm g, then divide by the moment of inertia Y to obtain the angular acceleration. Based on linear acceleration a and angular acceleration Combined with the construction of external perturbation vector and define control inputs To predict the desired descent rate, Model Predictive Control (MPC) employs a discrete state-space model to predict the evolution of the future state vector:

[0062] in, This is the state vector at the next control moment, where z is the control moment. It is the external disturbance vector obtained by wind load disturbance mapping, where A, B, and G are the state transition matrix, control input matrix, and disturbance input matrix, respectively. MPC is solved using finite-time optimization, with the objective function being:

[0063] Where S is the objective function, It is predicted at control time z. Current state Here, Q is the reference state, Q is the state deviation weight matrix, and R is the control quantity weight matrix. It is a slack variable. It is the penalty factor, and U is the prediction step size setting. It is predicted at control time z. The control input at each time step, where N is the total prediction step size; MPC optimization needs to satisfy input constraints (hard constraints), tilt angle constraints, and wind pressure soft constraints:

[0064] in, , These are the minimum and maximum control inputs, respectively. , These are the minimum and maximum tilt angles, respectively. That is the maximum wind pressure. yes Control input at control time, yes Control the tilt angle at any moment. yes Control the wind pressure at all times. yes Slack variables at control times; MPC solves a quadratic programming (QP) problem to optimize the objective function S, and uses a warm start (using the previous solution as the initial value) to improve efficiency. When the solution is successful, MPC outputs the optimal control input command every second. The actuator controls the descent of the equipment until it reaches the minimum height or receives a manual stop command. After the descent stops, the red "Do Not Lift" indicator light illuminates, triggering both mechanical and software double locking to prevent the equipment from lifting immediately. The locking status is reported to the remote platform, and the completion time is recorded.

[0065] By converting predicted wind pressure into an external disturbance vector using linear and angular acceleration, and defining full-state vectors for height, velocity, and attitude, accurate modeling of the equipment's wind-induced dynamics is achieved. Combined with model predictive control, look-ahead optimization is performed within a finite time domain. Weights and relaxation variables are introduced to balance velocity, stability, and energy consumption. Safety boundaries are ensured through input, tilt angle, and wind pressure constraints. Quadratic programming and warm start are used to improve real-time solution efficiency. Optimal commands are output every second to ensure a smooth descent. Upon completion, audible and visual alerts and dual locking are triggered, and the data is reported to a remote platform. Ultimately, the equipment is safe, controllable, responsive, and traceable.

[0066] Furthermore, when the environment meets safety conditions again, the lockout is lifted, and the controller's recovery indicator is used to assess the forecast for the future period. If the following conditions are met simultaneously: meteorological conditions (external meteorological or on-site monitoring confirms no rainfall or lightning warnings), control command conditions (the remote platform has not issued a "prohibit recovery" command, and the local lockout is lifted), and wind pressure prediction conditions (the maximum possible value of wind pressure at a significance level of 0), the lockout is lifted. Less than the critical wind pressure threshold When switching to "lift mode", the control input is the rising speed. During the lifting process, the wind pressure prediction conditions are continuously evaluated until the equipment is fully lifted into position.

[0067] By simultaneously verifying meteorological conditions, control commands, and wind pressure forecasts, a multi-source interlocking ascent trigger mechanism was implemented. This ensures that the ascent only starts within the safety boundary. During execution, the ascent speed is used as the control input, and wind pressure forecasts are continuously monitored to form a dynamic risk assessment and real-time interruption capability. This effectively prevents accidents midway through the process. At the same time, it takes into account both remote commands and local interlocks to avoid misoperation and conflicts. Ultimately, this achieves the goals of ensuring the safety of personnel and equipment, reducing false alarms and false starts, and improving automation and traceability.

[0068] This embodiment also provides a radar elevation control system based on meteorological monitoring, including: The power-on startup module is used to complete the power communication switch when the radar is powered on, load and verify the prediction model and control parameters, and establish a remote meteorological channel. The weather forecasting module is used to collect multi-element data in real time from the weather station, calculate the average wind speed using a sliding window, and combine XGBoost and EKF for fusion forecasting. The wind pressure assessment module is used to correct the mean and variance of wind speed based on a truncated normal distribution, derive the mean and variance of predicted wind pressure, and assess extreme risks. The risk warning module is used to trigger an alarm on-site by means of on-site sound and light devices when the wind pressure prediction exceeds the threshold to trigger a high risk, and to push the MPC control into the preparation state. The control and locking module is used for MPC to construct QP optimization solution for descent control input. After the device executes descent, it triggers double locking and records and reports the status.

[0069] This embodiment also provides a computer device applicable to a radar elevation control method based on meteorological monitoring, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the radar elevation control method based on meteorological monitoring as proposed in the above embodiment.

[0070] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0071] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a radar elevation control method based on meteorological monitoring as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

Claims

1. A radar elevation control method based on meteorological monitoring, characterized in that: include, After the radar is powered on, it completes the power and communication switching, initializes the controller, collects weather station data, and generates the future fused wind speed in real time through the Kalman filter physical model and residual neural network. The future wind speed is converted into wind pressure for assessment. Based on the assessment results, the risk level is determined and an early warning is issued. The controller is then preheated and put into the lifting preparation state. Based on the warning, the controller is activated to perform a descent operation and lock the equipment to prevent it from rising. When the environment meets the safety conditions again, the lock is released and the equipment is raised through the controller.

2. The radar elevation control method based on meteorological monitoring as described in claim 1, characterized in that: The collected meteorological station data is processed using a Kalman filter physical model and a residual neural network to generate a future fused wind speed index in real time. On-site weather data is collected from the split-type meteorological station at a fixed sampling frequency. The operating variables of the radar lifting equipment are collected through internal sensors. The instantaneous wind speed values ​​of the most recent N sampling times are stored in a circular buffer using a sliding window, and the average wind speed within the time window at each time moment is calculated to construct an average wind speed sequence. The first-order gradient feature is obtained by calculating the difference between instantaneous wind speeds at adjacent times. The input feature vector is constructed by combining meteorological data and operational variables as external influencing factors. The data is input into an XGBoost regression model for prediction. The mean wind speed prediction is calculated for each prediction step, and then compared with actual wind speed observations. Compared with the average wind speed forecast Subtraction calculation of predicted residuals A residual circular buffer is established, and the variance of the predicted residual is used. As an estimate of the prediction uncertainty, the final output prediction pair ; Retrieve the mean wind speed forecast at discrete times from a remote API. The prediction variance is calculated by subtracting the historical API wind speed forecast mean from the actual measured value. Furthermore, piecewise cubic Hermite interpolation with monotonically preserved parameters is used to transform discrete-time meteorological forecast data into second-by-second sequences, and the prediction variance is analyzed. Perform dilation correction to obtain the interpolated corrected remote API variance. Assuming the remote API data is available, output the interpolated remote API prediction. Furthermore, an adaptive extended Kalman filter method is used to recursively estimate the wind speed state, defining a state vector. The state equations are in the form of nonlinear damping and process noise is incorporated, based on the state vector. The observation equation is defined, and the filtering recursion follows a two-stage process of "prediction-update". In the prediction stage, the prior state estimate vector at the current time is calculated through a nonlinear state transition function. The prior error covariance matrix is ​​obtained by propagating the covariance based on the Jacobian matrix. Based on the prior error covariance matrix and observed values Calculate the Kalman gain to obtain the Kalman gain matrix. The posterior state estimate vector is obtained by correcting the prior state estimate vector and the prior error covariance matrix at the current time. and posterior error covariance matrix , as the initial condition for prediction at the next moment; In the EKF update step, the measured wind speed at time k is used. With predicted wind speed The observation residuals are obtained by subtraction. Perform measurement noise covariance update based on observed residuals. and the updated measurement noise covariance Correct the noise covariance during the process and construct the input vector of the neural network. , input vector Input to neural network Output state correction and the posterior state estimation vector The final state estimate is obtained by summing the residuals. Estimate based on the final state As the initial state, the nonlinear state transition function Continuous stacking Step by step, the predicted state at each step is iterated by substituting the predicted state of the previous step into the nonlinear state transition function to obtain the future state. Step state prediction vector Simultaneously, the prediction error covariance matrix is ​​recursively derived. Combined with future state prediction vector Output predicted wind speed and variance. The XGBoost regression model prediction results are weighted with the remote API prediction results to obtain the fusion model's wind speed prediction value and variance. When the remote API is unavailable, it degenerates into a single model, and the model forecast results are displayed. The predicted wind speed is then fused with the prediction results from the adaptive extended Kalman filter method using a second weighted fusion method. variance is .

3. The radar elevation control method based on meteorological monitoring as described in claim 2, characterized in that: The process of converting future wind speed into wind pressure for evaluation refers to, since the actual wind speed is greater than or equal to 0, correcting the wind speed distribution parameters based on a left-truncated normal distribution, calculating the standardized cutoff point Z, and setting a threshold. When the standardized cutoff point Z is less than the threshold If the truncation effect is negligible, no correction is needed; otherwise, parameter correction is performed. For cases requiring correction, the standard normal density function is calculated. and cumulative distribution function Based on the existing derivation formula for truncated normal moments, the mean and variance of wind speed are corrected, and the one-sided confidence quantile at a given significance level 0 is calculated using the inverse function of the monotonic approximation function. The mean predicted wind pressure for each prediction step is calculated based on the corrected mean and variance of wind speed. Based on the corrected variance The Delta method is used to propagate the uncertainty in wind speed prediction to wind pressure, thus obtaining the variance in wind pressure prediction. Variance prediction by wind pressure Define wind pressure standard deviation By one-sided confidence quantiles Calculate the one-sided confidence maximum value of wind pressure at a significance level of o. Set the critical wind pressure threshold The maximum value of wind pressure at the significance level o Greater than or equal to the critical wind pressure threshold When this happens, it is determined to enter a "high-risk" state, and debouncing logic is used, that is, within the range If two or more prediction points exceed the threshold consecutively, the risk level is immediately confirmed as "high risk"; otherwise, the prediction cycle will be repeated for verification.

4. The radar elevation control method based on meteorological monitoring as described in claim 3, characterized in that: The step of issuing an early warning based on the assessment results and preheating the controller to enter the lifting preparation state refers to the detection of the predicted step size. When the risk level is high, the on-site audible and visual warning device should be activated immediately to provide sound warnings, optical warnings, and human-machine interface prompts. The warning information should be packaged into a standardized data format and sent to the remote monitoring platform. At the same time as issuing the warning, the MPC control should be put into the ready state.

5. The radar elevation control method based on meteorological monitoring as described in claim 4, characterized in that: The start controller performs a descent operation and locks the device to prevent it from rising. It also collects and confirms the physical parameters of the device and its mounting components. At each control moment, the state vector is defined as follows: And by using the predicted wind pressure average Multiply by the equivalent windward area O to obtain the equivalent wind load. Equivalent wind load Dividing by the mass m to obtain the linear acceleration a, the equivalent wind load is then converted. Multiply by the torque arm g, then divide by the moment of inertia Y to obtain the angular acceleration. Based on linear acceleration a and angular acceleration Combined with the construction of external perturbation vector and define control inputs To achieve the desired descent rate, Model Predictive Control (MPC) employs a discrete state-space model to predict the future state vector evolution. MPC solves this problem using a finite-time optimization approach, with an objective function S. The MPC optimization must satisfy input constraints, tilt angle constraints, and soft wind pressure constraints. MPC solves a first-order quadratic programming problem to optimize the objective function S. Upon successful solution, MPC outputs the optimal control input command per second. During the descent process, the equipment descends until it reaches the minimum height or receives a manual stop command. Once the descent stops, the red "Do Not Lift" indicator light illuminates, triggering both mechanical and software locking to prevent the equipment from lifting immediately. The locking status is then reported to the remote platform, and the completion time is recorded.

6. The radar elevation control method based on meteorological monitoring as described in claim 5, characterized in that: When the environment meets the safety conditions again, the lock is released and the forecast for the future period is evaluated by the controller. If the meteorological conditions, control command conditions, and wind pressure forecast conditions are met at the same time, the device is switched to "lift mode". At this time, the control input is the lifting speed. During the lifting process, the wind pressure forecast conditions are continuously evaluated until the device is fully lifted into place.

7. The radar elevation control method based on meteorological monitoring as described in claim 6, characterized in that: After the radar is powered on, it completes the power and communication switching and initializes the controller. After the radar system is powered on, it detects the power status and establishes a communication link. After the power and communication are stable, it starts the core hardware and loads the required models and parameters. During the loading process, it verifies the version, hash value and size of the file and establishes a remote meteorological API channel to receive weather forecast data. When the remote meteorological API channel is unavailable, it enables the local micro-area wind field extrapolation module and records the degradation status flag.

8. A radar elevation control system based on meteorological monitoring, based on the radar elevation control method based on meteorological monitoring as described in any one of claims 1 to 7, characterized in that: include, The power-on startup module is used to complete the power communication switch when the radar is powered on, load and verify the prediction model and control parameters, and establish a remote meteorological channel. The weather forecasting module is used to collect multi-element data in real time from the weather station, calculate the average wind speed using a sliding window, and combine XGBoost and EKF for fusion forecasting. The wind pressure assessment module is used to correct the mean and variance of wind speed based on a truncated normal distribution, derive the mean and variance of predicted wind pressure, and assess extreme risks. The risk warning module is used to trigger an alarm on-site by means of on-site sound and light devices when the wind pressure prediction exceeds the threshold to trigger a high risk, and to push the MPC control into the preparation state. The control and locking module is used for MPC to construct QP optimization solution for descent control input. After the device executes descent, it triggers double locking and records and reports the status.

9. A computer device, comprising: Memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, it implements the steps of the radar elevation control method based on meteorological monitoring as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the radar elevation control method based on meteorological monitoring as described in any one of claims 1 to 7.

Images