Automatic fog dispersal control method and system based on microphysical evaluation model of warm fog
By adopting an automatic spraying control method based on a microphysical evaluation model, the problems of relying on manual experience and difficulty in balancing energy consumption in the control of warm fog dispersing devices have been solved, thus achieving efficient, stable and economical operation of fog dispersing operations.
Patent Information
- Application Number
- CN202611595944.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-15
- Publication Date
- 2026-11-10
AI Technical Summary
Existing warm fog defogging devices rely on manual experience for control, making it difficult to achieve stability in defogging effect and balance energy consumption. Furthermore, they lack closed-loop feedback and adaptive adjustment capabilities, making it difficult to operate efficiently in complex and ever-changing fog environments.
An automatic seeding control method based on a microphysical evaluation model is adopted. By receiving fog dissipation task parameters and real-time acquisition of fog field environmental parameters, the Kunkel or Reuge mode is adaptively called for microphysical evaluation. The objective function is constructed and the four-dimensional control parameters are solved using a multivariate gradient descent optimization algorithm, forming a closed-loop feedback control system.
It enables proactive prediction of fog control, improves the automation level and stability of fog control operations, reduces the amount of ineffective catalyst application, lowers the energy consumption of the unit, and improves the accuracy and response speed of parameter adjustment.
Smart Images

Figure CN123067084A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of warm fog dissipation operations, and in particular to an automatic spraying control method and system for warm fog dissipation based on a microphysical evaluation model. Background Technology
[0002] Fog is defined as a near-ground horizontal visibility of less than 1 km, primarily formed by tiny water droplets suspended in the air. Typical radiation fog typically has a fog layer height between 100 and 200 meters, and its formation is closely related to radiative cooling and turbulent mixing processes within a stable boundary layer. In recent years, frequent occurrences of low-visibility weather such as dense fog have posed serious safety threats to civil aviation takeoffs and landings, port and shipping operations, road traffic, and the security of major events, creating an urgent need for efficient, precise, and rapidly deployable active fog dispersal technologies.
[0003] Hygroscopic solid powder catalysts are the mainstream catalytic method in the field of warm fog defogging. The defogging process goes through four stages in sequence: rapid deliquescence, water vapor evaporation-condensation, particle collision and growth, and sedimentation defogging. The physical properties, particle size distribution, and spreading control precision of the catalyst directly determine the efficiency and cost of the final defogging operation. With the continuous expansion of defogging operation scenarios, higher requirements are placed on the scientific, intelligent, and refined level of the spreading control system.
[0004] Currently, conventional warm fog dispersal and spraying control methods mainly rely on operators' on-site experience or preset thresholds to set operating parameters such as the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed of the dispersal device. The gimbal attitude control, feed output, and fan spraying systems of the dispersal device mostly adopt independent control modes, and no collaborative optimization mechanism based on the dispersal effect is established between the control commands of each actuator. This not only makes it difficult to guarantee the stability of the dispersal effect but also easily leads to ineffective catalyst consumption and wasted energy. Furthermore, the entire control process lacks closed-loop feedback and adaptive adjustment capabilities, resulting in low automation, delayed control response, and difficulty in achieving a balance between dispersal efficiency and operating costs in complex and variable fog environments. Summary of the Invention
[0005] To achieve a balance between fog dispersal efficiency and operating costs, this application provides an automatic spraying control method and system for warm fog dispersal based on a microphysical evaluation model.
[0006] Firstly, this application provides an automatic spraying control method for warm fog dispersal based on a microphysical evaluation model, employing the following technical solution: An automatic spraying control method for warm fog dispersal based on a microphysical evaluation model includes: Receive fog removal task parameters, and acquire fog field environmental parameters, catalyst physical property parameters, and actuator status parameters of the fog removal device in real time, and verify the validity of the parameters; Based on the verified fog field environmental parameters, the fog field type is determined, and a microphysical evaluation model is established by adaptively calling the Kunkel mode or Reuge mode. The evolution parameters of the fog field microphysical process are calculated in real time, and based on the evolution parameters of the fog field microphysical process, the predicted values of horizontal visibility and fog layer elimination rate at future times are output. The predicted horizontal visibility value is compared with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, an objective function containing fog dissipation efficiency and time consumption cost is constructed based on the microphysical evaluation model, and the multivariate gradient descent optimization algorithm is called to solve for the minimum value of the objective function, so as to obtain the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. Based on the incremental generation of the four-dimensional control parameters, control commands are generated to drive the fog dispersing device to perform the spreading action, and the measured data after execution is collected and fed back to the microphysical evaluation model as the initial conditions for the next iteration.
[0007] By adopting the above technical solution, automatic spraying control for warm fog dispersal is achieved based on a microphysical evaluation model. A complete automated control system is formed, from data input, model construction, parameter optimization to closed-loop feedback. This effectively solves the technical problems of traditional warm fog dispersal operations, such as reliance on manual experience for spraying parameters, poor fog field adaptability, and difficulty in balancing dispersal efficiency and energy consumption. Through multi-dimensional real-time acquisition and validity verification of dispersal task parameters, fog field environmental parameters, catalyst property parameters, and actuator status parameters, all input conditions for dispersal operations can be comprehensively covered. Based on this, a microphysical evaluation model is established by adaptively calling either the Kunkel or Reuge mode according to the fog field type. This allows for matching a more adaptable calculation mode to the microphysical structure and evolution laws of different warm fog fields, overcoming the limitation of insufficient generalization ability of a single fog microphysical model. The real-time calculated fog field microphysical process evolution parameters and predicted values of future horizontal visibility and fog layer elimination rate can accurately predict the fog field development trend and dispersal operation effect, realizing a shift from passive response to proactive prediction in fog dispersal control. When the predicted visibility fails to meet the target, an objective function integrating fog dispersal efficiency and time consumption is constructed. A multivariate gradient descent optimization algorithm is then used to solve for the incremental four-dimensional control parameters of the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed. This allows for a balanced approach to operational efficiency and cost while meeting fog dispersal requirements, achieving optimal configuration of parameters across multiple actuators. This replaces the traditional manual trial-and-error parameter tuning method, improving the accuracy and response speed of parameter adjustments. Finally, an iterative mechanism based on feedback from measured data after seeding to the microphysical evaluation model forms a closed-loop control chain of "prediction-optimization-execution-feedback." This continuously corrects the model's calculation boundaries based on real fog field feedback, ensuring the microphysical evaluation model's output constantly aligns with the actual fog field evolution characteristics. This drives continuous dynamic optimization of control parameters, improving the overall automation level, fog dispersal target achievement rate, and operational stability of warm fog dispersal operations. Simultaneously, it effectively reduces the amount of ineffective catalyst seeding, lowering energy consumption and operating costs.
[0008] Optionally, the steps of determining the fog field type based on the verified fog field environmental parameters, adaptively calling the Kunkel or Reuge mode to establish a microphysical evaluation model, and calculating the evolution parameters of the fog field microphysical processes in real time specifically include: Based on the aforementioned fog field environmental parameters, three indices—turbulent dissipation rate, vertical wind speed root mean square error, and fog droplet spectrum fitting goodness of fit—are jointly determined for evaluation. When the turbulent dissipation rate is less than the first threshold, the vertical wind speed root mean square error is less than the second threshold, and the goodness of fit of the fog droplet spectrum is greater than the third threshold, the fog field is determined to be a homogeneous weak turbulent state, and the Kunkel model is called to calculate the water vapor consumption. If any of the criteria is not met, the fog field is determined to be in a heterogeneous turbulent state, and the process is switched to Reuge mode to solve the particle number density balance equation and the gas-liquid mass balance equation.
[0009] By adopting the above technical solutions and the on-demand intelligent mode switching mechanism, the real-time computing efficiency of the system under stable fog fields is guaranteed, while the microphysical simulation accuracy under complex turbulent fog fields is also guaranteed. This achieves a dynamic balance between computing efficiency and simulation accuracy, effectively expands the applicability of the automatic seeding control system to different types of warm fog fields, and further improves the stability of the actual operation effect of fog dispersal.
[0010] Optionally, the step of outputting predicted horizontal visibility and predicted fog dissipation rate at future times based on the evolution parameters of the fog field microphysical processes specifically includes: When Kunkel mode is invoked, the feed mass flow rate, dispersion efficiency, and effective diffusion volume are determined based on the actuator state parameters. Combined with the particle mass in the catalyst physical property parameters, the catalyst number density per unit volume is calculated. The mass change during particle descent is calculated based on the water vapor diffusion equation. The water vapor consumption per unit volume is obtained by combining the catalyst number density and the mass change. The water vapor consumption per unit volume is then substituted into the visibility diagnostic equation to output the horizontal visibility prediction value and the fog elimination rate prediction value. or, When the Reuge mode is invoked, the particle number density balance equation and the gas-liquid mass balance equation, which include the turbulence dissipation coefficient and the collision efficiency factor, are solved to obtain the updated droplet size distribution and liquid water content. The updated liquid water content is then substituted into the visibility diagnostic equation to output the predicted horizontal visibility and the predicted fog layer elimination rate.
[0011] By adopting the above technical solutions, both calculation modes have established a closed-loop quantitative mapping relationship of "spreading execution parameters - fog field microphysical parameters - fog elimination effect indicators" and uniformly use the visibility diagnostic equation as the final output interface. This not only ensures the rationality and accuracy of the calculation in their respective applicable scenarios, but also achieves the uniformity of the prediction output under different modes.
[0012] Optionally, the steps of using the multivariate gradient descent optimization algorithm to solve for the minimum of the objective function and obtain the four-dimensional control parameter increments specifically include: Initialize the parameters of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed as the current optimization parameters; Based on the current optimization parameters, the visibility changes in future time periods are predicted using the microphysical evaluation model, and a first visibility time series curve is generated. In parallel, while keeping the optimized parameter values from the previous cycle unchanged, the visibility changes in the corresponding future time period are predicted to generate a second visibility time series curve. Calculate the difference between the first visibility time series curve and the second visibility time series curve. When the marginal improvement rate of the difference is lower than a preset efficiency threshold, it is determined that the efficiency boundary has been reached and the feeder speed increment is limited. The gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed parameters that minimize the objective function are obtained by iteratively solving the rule update through gradient descent. The updated current optimization parameters are then used to generate the next round of visibility time series curves until the convergence condition is met.
[0013] By adopting the above technical solution, and by constructing a dual-dimensional objective function that takes into account both fog elimination efficiency and time cost, and combining an integrated optimization architecture of visibility trajectory difference evaluation, marginal efficiency boundary constraints, and multivariable gradient descent iterative solution, the coordinated optimization of four-dimensional control parameters—gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed—was achieved. Under the premise of ensuring that the fog elimination effect meets the standards, the operation timeliness and operating cost were effectively balanced, and the rationality, economy, and convergence efficiency of the seeding parameter optimization were improved.
[0014] Optionally, the step of iteratively solving for the four-dimensional control parameters using gradient descent to update the rule further includes: The target fog layer height is determined based on the fog top height in the fog field environmental parameters. Combined with the horizontal coverage range in the fog dissipation task parameters, the optimal combination of gimbal elevation angle and fan speed is solved by using a particle ballistic model and combining the particle size and drag coefficient in the catalyst physical property parameters. The lateral drift during particle flight is calculated based on the horizontal wind speed and wind direction deviation in the fog field environmental parameters, and a feedforward compensation is superimposed in the gimbal azimuth control loop to counteract the drift effect.
[0015] By adopting the above technical solution, the physical mechanism model and feedforward compensation are integrated into the design of the gradient descent optimization system, realizing the deep integration of numerical optimization and physical laws. It not only retains the global optimization capability of the multivariable gradient descent algorithm for the collaborative optimization of parameters of multiple actuators, but also makes up for the shortcomings of pure algorithm optimization in terms of spatial accuracy and environmental adaptability through the mechanism analysis of particle motion. This makes the final output parameters of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed both globally optimal and reasonable for on-site operation.
[0016] Optionally, the step of iteratively solving for the four-dimensional control parameters using gradient descent to update the rule further includes: Set physical boundary constraints for the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed; The gradient descent iterative calculation results are forcibly restricted to the range of physical boundary constraints to ensure that the control instructions are within the executable range.
[0017] By adopting the above technical solutions, the boundary constraint mechanism, together with the aforementioned objective function optimization, particle ballistic model inverse solution, and wind-induced drift feedforward compensation, forms a hierarchical and collaborative optimization architecture. While ensuring the global optimality, spatial accuracy, and environmental adaptability of parameters, it firmly adheres to the bottom line of engineering feasibility, achieving a deep match between the algorithm optimization effect and the actual hardware capabilities, and further enhancing the engineering practicality and operational reliability of the entire warm fog dissipation automatic seeding control method.
[0018] Optionally, after the step of collecting and feeding back the measured data after execution to the microphysics evaluation model as the initial condition for the next iteration, the method further includes: Obtain the measured visibility sequence and liquid water content sequence; Determine the current call mode type: If the current mode is Kunkel, then the least squares fitting algorithm is used to dynamically correct the dispersion efficiency and effective diffusion volume parameters in the microphysical evaluation model based on the measured visibility sequence and the liquid water content sequence. If the current mode is Reuge, then the least squares fitting algorithm is used to dynamically correct the turbulence dissipation coefficient and collision efficiency factor in the microphysical evaluation model based on the measured visibility sequence and the liquid water content sequence. The corrected parameters are then applied to subsequent calculations.
[0019] By adopting the above technical solution, a dynamic correction mechanism for model coefficients based on measured data is introduced under the closed-loop feedback iterative control framework. This can effectively solve the problem of mismatch between empirical set parameters and real-time characteristics of actual fog fields in the microphysics evaluation model, suppress the error accumulation effect in the multi-round iteration process, continuously improve the model's calculation accuracy and fog field adaptability, and provide self-calibration capability support for the long-term stable operation of the entire automatic seeding control system.
[0020] Optionally, the step of comparing the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters further includes: When the predicted horizontal visibility value is less than the preset target threshold and the cumulative operation time exceeds the maximum allowable time, an emergency response strategy is triggered. The feeder speed and fan speed are forcibly set to their respective upper limits to perform extreme fogging intervention.
[0021] By adopting the above technical solutions, this emergency support mechanism complements the core modules such as adaptive construction of microphysical models, optimization of multivariable gradient parameters, and closed-loop feedback dynamic calibration, forming a complete control architecture of "normal fine-grained control - extreme condition backup". This further improves the full-scenario adaptability of the automatic warm fog dissipation control method, ensuring both efficient and economical operation of routine work and providing a solid safety barrier for achieving tasks under extreme conditions.
[0022] Optionally, the steps for validating the parameters may include: The physical reasonable threshold range of the fog field environmental parameters, the catalyst physical property parameters, and the actuator status parameters is verified. When parameters exceed limits, an alarm is triggered and the system is prevented from starting. After verification, the ambient temperature, air pressure, liquid water content, and wind field parameters are standardized and packaged as input boundary conditions for the microphysical evaluation model.
[0023] Secondly, this application provides an automatic warm fog dispersal control system based on a microphysical evaluation model, which adopts the following technical solution: An automatic spraying control system for warm fog dispersal based on a microphysical evaluation model includes: The data acquisition module is used to receive fog removal task parameters, acquire fog field environmental parameters, catalyst physical property parameters and actuator status parameters of the fog removal device in real time, and verify the validity of the parameters. The microphysics assessment engine module is used to determine the fog field type based on the fog field environmental parameters after verification, adaptively call the Kunkel mode or Reuge mode to build a microphysics assessment model, calculate the evolution parameters of the fog field microphysics process in real time, and output the predicted values of horizontal visibility and fog layer elimination rate at future times based on the fog field microphysics process evolution parameters. The decision control module is used to compare the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, it constructs an objective function containing fog dissipation efficiency and time consumption cost based on the microphysical evaluation model, and calls the multivariate gradient descent optimization algorithm to solve for the minimum value of the objective function, thereby obtaining the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. The execution drive module is used to generate control commands based on the incremental four-dimensional control parameters to drive the defogging device to perform the spreading action. The feedback calibration module is used to collect the measured data after execution and feed it back to the microphysics evaluation engine module as the initial condition for the next iteration.
[0024] In summary, this application includes at least one of the following beneficial technical effects: 1. Adaptive microphysical simulation and advanced prediction capabilities for fog fields: The fog field state is determined by jointly using three indicators: turbulence dissipation rate, vertical wind speed mean square error, and fog droplet spectrum fitting goodness of fit. The microphysical evaluation model is constructed by adaptively matching the Kunkel model or the Reuge model, which takes into account both the computational efficiency of homogeneous weak turbulent fog fields and the simulation accuracy of heterogeneous turbulent fog fields. It can quantitatively output the predicted values of horizontal visibility and fog layer elimination rate at future times, realizing the transformation of fog control from passive response to active prediction.
[0025] 2. Highly efficient and economical seeding control through multi-parameter collaborative optimization: A dual-dimensional objective function integrating fog-eliminating efficiency and time consumption cost is constructed. The incremental control parameters of four dimensions—gimbal elevation angle, azimuth angle, feeder speed, and fan speed—are solved using a multivariate gradient descent algorithm. Simultaneously, the inverse solution of the particle ballistic model, wind-induced drift feedforward compensation, and physical boundary constraint mechanisms are embedded to achieve the optimal collaborative configuration of multiple actuators. This replaces the traditional manual trial-and-error parameter adjustment method, effectively reducing the amount of ineffective catalyst seeding and the energy consumption of the device while ensuring that the fog-eliminating effect meets the standards, thereby improving the accuracy of parameter adjustment and the economic efficiency of operation.
[0026] 3. Closed-loop self-calibration and hierarchical fallback for stable operation across all scenarios: Based on feedback from measured data, an iterative closed loop is formed. The turbulence dissipation coefficient and collision efficiency correction factor of the model are dynamically corrected through least squares fitting, which suppresses the accumulation of errors from multiple iterations and continuously improves the model's adaptability to real-time fog fields. At the same time, an overdue emergency response strategy is added to form a hierarchical control architecture of "normal fine optimization + extreme intervention under extreme conditions". This ensures both the economical operation of routine work and the adherence to the timeliness of tasks, and comprehensively improves the system's adaptability to complex and ever-changing fog conditions and operational reliability. Attached Figure Description
[0027] Figure 1 This is a first flowchart of an embodiment of the method of this application; Figure 2 This is a second flowchart of an embodiment of the method of this application; Figure 3 This is a structural block diagram of an embodiment of the system in this application. Detailed Implementation
[0028] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-3 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.
[0029] The first embodiment of this application discloses an automatic spraying control method for warm fog dispersal based on a microphysical evaluation model. (Refer to...) Figure 1The automatic spraying control method for warm fog dissipation includes S110-S140: S110 receives fog removal task parameters, and acquires fog field environmental parameters, catalyst physical property parameters, and actuator status parameters of the fog removal device in real time, and performs legality verification on the parameters; S120 determines the fog field type based on the verified fog field environmental parameters, adaptively calls Kunkel mode or Reuge mode to establish a microphysical evaluation model, calculates the evolution parameters of fog field microphysical processes in real time, and outputs the predicted values of horizontal visibility and fog layer elimination rate for future times based on the evolution parameters of fog field microphysical processes. S130 compares the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, it constructs an objective function containing fog dissipation efficiency and time consumption cost based on the microphysical evaluation model, and calls the multivariate gradient descent optimization algorithm to solve for the minimum value of the objective function, thereby obtaining the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. S140 generates control commands based on the incremental generation of four-dimensional control parameters to drive the fog dispersing device to perform the spreading action, and collects the measured data after execution and feeds it back to the microphysical evaluation model as the initial conditions for the next iteration.
[0030] Specifically, step S110 can be achieved by using an industrial-grade industrial control computer with multiple types of sensor acquisition links. After the system starts up, it first receives the fog removal task parameters entered by the operator through the human-machine interface, including the target visibility threshold Vtarget, the maximum allowable operation time Tmax, the target fog area horizontal coverage Ltarget, and the specified catalyst model.
[0031] Meanwhile, the system connects in real time via Modbus to the front-end forward scattering visibility meter, droplet spectrometer, 3D ultrasonic anemometer, integrated temperature, humidity and air pressure sensor, and the device's own status encoder. It simultaneously acquires fog field environmental parameters, catalyst property parameters, and the status parameters of the defogging device's actuators at a sampling frequency of 1Hz. The fog field environmental parameters include ambient temperature T, atmospheric pressure P, ambient water vapor pressure e, fog layer thickness H, fog top height Htop, and initial liquid water content. Average droplet radius ra, droplet number density Na, turbulent dissipation rate ε, horizontal wind speed U, vertical wind speed W, and wind direction deviation angle. The catalyst's physical properties are retrieved from the built-in physical property parameter library based on the entered model, including solute molecular weight Ms and catalyst density. molar ion number v, osmotic pressure coefficient Median particle size Geometric standard deviation The temperature-dependent solubility Ssol(T) is used; the actuator status parameters are directly read from the feedback registers of the gimbal encoder and the frequency converter driver, including the current elevation angle θ and azimuth angle. The feeder speed ωf and the fan speed ωb.
[0032] In S110, the steps for validating the parameters specifically include: Verify the physical reasonable threshold range of fog field environmental parameters, catalyst physical property parameters, and actuator status parameters; When parameters exceed limits, an alarm is triggered and the system is prevented from starting. After verification, the ambient temperature, air pressure, liquid water content, and wind field parameters are standardized and packaged as input boundary conditions for the microphysics evaluation model.
[0033] Specifically, the parameter validity verification process is completed by a built-in threshold verification subroutine. The system presets a physically reasonable threshold range for each type of parameter. For example, the ambient temperature verification range is set to 0~30℃, the atmospheric pressure verification range is 950~1050hPa, the gimbal elevation angle verification range is 0°~90°, and the feeder and fan speed verification range is 0 to the rated speed of the equipment. When any parameter exceeds the threshold range, the system immediately triggers a red audible and visual alarm on the industrial control interface and simultaneously locks the start circuit of the actuator, preventing the system from entering the working state.
[0034] After successful verification, the system performs standardized encapsulation and unit conversion on all input parameters. Temperature is uniformly converted to thermodynamic Kelvin (T_K = T_℃ + 273.15), liquid water content is uniformly expressed as g / m³, wind speed as m / s, internal gradient and trigonometric function calculations use radians, while the interface display retains angles. The saturated vapor pressure of pure water is calculated in real-time using the Magnus formula. The calculation results are in kPa and are finally packaged into three structured data packets: environmental parameter set, catalyst parameter set, and actuator state parameter set. These packets are transmitted to the input buffer of the microphysics evaluation engine via a real-time Ethernet bus as input boundary conditions for subsequent model calculations.
[0035] In S120, the steps of determining the fog field type based on the verified fog field environmental parameters, adaptively calling Kunkel or Reuge mode to establish a microphysical evaluation model, and calculating the evolution parameters of the fog field microphysical processes in real time specifically include: The three indices of turbulent dissipation rate, vertical wind speed root mean square error, and droplet spectrum fitting goodness of fit are determined based on the fog field environmental parameters and then jointly judged. When the turbulent dissipation rate is less than the first threshold, the vertical wind speed root mean square error is less than the second threshold, and the goodness of fit of the fog droplet spectrum is greater than the third threshold, the fog field is determined to be a homogeneous weak turbulent state, and the Kunkel model is called to calculate the water vapor consumption. If any of the criteria is not met, the fog field is determined to be in a heterogeneous turbulent state, and the process is switched to Reuge mode to solve the particle number density balance equation and the gas-liquid mass balance equation.
[0036] Specifically, in step S120, after completing parameter reception, verification, and encapsulation, the system enters the fog field type determination and mode adaptive switching stage. This stage is implemented by a built-in dynamic switching arbitrator. The arbitrator first extracts the turbulence dissipation rate, the vertical wind speed time series over the past 10 seconds, and the measured fog droplet size distribution data from the standardized environmental parameter set. It then calculates the standard deviation of the vertical wind speed samples to obtain the root mean square error of the vertical wind speed. The calculation formula is as follows: In the formula, N is the number of samples, and Wmean is the average vertical wind speed during the period. Then, the log-normal distribution function is used to perform least squares fitting on the measured fog droplet spectrum to obtain the fitted scale distribution function ffit(r), and the goodness of fit determination coefficient R² is calculated. The formula is R²=1-Σ(fobs(r)-ffit(r))² / Σ(fobs(r)-fmean)², where fobs(r) is the measured fog droplet number density distribution, and fmean is the average value of the measured distribution.
[0037] Subsequently, a joint judgment is performed based on three indicators. The first threshold for turbulent dissipation rate is set at 0.01 m² / s³, the second threshold for the root mean square error of vertical wind speed is set at 0.05 m / s, and the third threshold for the goodness of fit of the fog droplet spectrum is set at 0.95. When all three indicators simultaneously satisfy ε≤0.01 m² / s³, σW≤0.05 m / s, and R²≥0.95, the current fog field is determined to be a homogeneous weak turbulent state, and the system calls the Kunkel mode as the evaluation kernel to ensure computational efficiency. When any one of the judgment conditions is not met, the fog field is determined to be a heterogeneous turbulent state, and the system immediately switches to the Reuge mode to ensure evaluation accuracy.
[0038] To avoid frequent mode switching caused by minor fluctuations in the fog field, the arbitrator adopts a delayed switching strategy. Switching from Reuge mode to Kunkel mode requires three consecutive calculation cycles to meet the homogeneity condition, while switching from Kunkel mode to Reuge mode can be executed immediately as long as the condition is not met in a single cycle. During the switching process, the liquid water content and visibility at the previous moment are retained as the initial calculation values to ensure smooth output without jumps. The entire judgment and switching process is completed within a single 1-second calculation step.
[0039] In S120, the steps for outputting the predicted horizontal visibility and fog dissipation rate for future times specifically include: When the Kunkel mode is invoked, the feed mass flow rate, dispersion efficiency, and effective diffusion volume are determined based on the actuator state parameters. Combined with the particle mass in the catalyst physical property parameters, the catalyst number density per unit volume is calculated. The mass change during particle descent is calculated based on the water vapor diffusion equation. The water vapor consumption per unit volume is obtained by combining the catalyst number density and the mass change. The water vapor consumption per unit volume is then substituted into the visibility diagnostic equation to output the horizontal visibility prediction value and the fog elimination rate prediction value. or, When the Reuge mode is invoked, the particle number density balance equation and the gas-liquid mass balance equation, which include the turbulence dissipation coefficient and the collision efficiency factor, are solved to obtain the updated droplet size distribution and liquid water content. The updated liquid water content is then substituted into the visibility diagnostic equation to output the predicted horizontal visibility and the predicted fog layer elimination rate.
[0040] Specifically, after entering the evaluation and calculation phase, the system executes the corresponding solution logic according to the mode selected by the arbitrator. When the Kunkel mode is invoked, the system first calculates the feed mass flow rate based on the actuator state parameters, using the following formula: In the formula, kf is the feeder speed-mass flow rate proportionality coefficient obtained through prior calibration tests, with units of... ωf is the current feeder speed.
[0041] Then, based on the current fan speed ωb, the dispersion efficiency curve calibrated in advance through cold spraying test is consulted to obtain the catalyst dispersion efficiency ηdisp(ωb) at the corresponding speed, with a value range of 0~1. At the same time, the effective diffusion volume Veff(θ,ωb) is calculated by a binary function of elevation angle and fan speed. This function is generated by a three-dimensional lookup table pre-calculated by the particle ballistic model, and the effective diffusion volume under any combination of elevation angle and speed can be quickly obtained by bilinear interpolation.
[0042] Subsequently, the catalyst density ρp and the initial particle radius were combined. Calculate the mass of a single catalyst particle (i.e., half the median particle size). The mass of a single particle is... Next, the average residence time of particles within the fog layer was introduced. This parameter is determined by the ratio of the target fog layer height Htarget to the final velocity vp of the falling particles, i.e., τres = Htarget / vp, and is used to characterize the residence time of catalyst particles in the effective diffusion space. Finally, based on the principle of steady-state mass conservation, the particle supply rate is obtained by dividing the effective mass flow rate by the mass of a single particle, multiplying it by the average residence time to obtain the total number of particles residing in the space, and then dividing it by the effective diffusion volume to obtain the catalyst number density per unit volume. The complete formula is then modified as follows: The physical meaning of this formula is the number of catalyst particles effectively dispersed within a unit volume of fog zone under steady-state dispersal conditions.
[0043] After completing the number density calculation, the system calls the water vapor diffusion equation to calculate the mass change ΔM of a single particle within a 1-second step Δt, using the following formula: In the formula, Rp(t) is the real-time radius of the particle at time t, and Dv is the water vapor diffusion coefficient, which is taken as 2.5 × 10⁻⁶ under normal temperature and pressure. -5 m² / s, e(t) is the ambient water vapor pressure, esol(t) is the saturated water vapor pressure at the surface of the solution droplet, and Rv is the water vapor constant, with a value of [value missing]. T_K is the ambient thermodynamic temperature.
[0044] The saturated vapor pressure on the surface of the solution droplet is calculated using the osmotic pressure correction equation, and the formula is as follows: In the formula, es(T_K) is the saturated vapor pressure of pure water at the same temperature. ν is the osmotic pressure coefficient, ν is the number of ions produced by the ionization of the solute, ms is the mass of the solute, Ms is the molar mass of the solute, mw is the mass of the solvent water in the solution, and Mw is the molar mass of water, with a value of 18 g / mol. This corrected formula quantitatively describes the effect of hygroscopic solutes reducing the equilibrium water vapor pressure on the droplet surface, and is the core physical basis for catalyst condensation and growth.
[0045] Multiplying the change in mass of a single particle by the catalyst number density per unit volume yields the water vapor consumption per unit volume. This corresponds to the reduction in liquid water content in the fog area. Finally, the updated liquid water content LWC(t+Δt)=LWC(t)-Δe is substituted into the visibility diagnostic equation, the formula is: In the formula, 'a' is an empirical coefficient related to the droplet spectrum, which can be dynamically adjusted through on-site calibration, and 'b' takes a default value of 0.7. The predicted horizontal visibility value V(t+Δt) for the next moment is calculated, and the predicted fog removal rate is calculated based on the ratio of the initial liquid water content to the current liquid water content. .
[0046] In the calculation, the final velocity vp of the falling particle is calculated using Stokes' theorem, which is as follows: This velocity is not only used to correct the residence time of particles within the fog layer, but is also a key parameter for calculating the average residence time τres; in the formula, g is the gravitational acceleration, taken as 9.81 m / s², dp is the real-time particle diameter, and μair is the aerodynamic viscosity, taken as at room temperature. It is used to correct the residence time and effective height of particles within the fog layer.
[0047] When Reuge mode is invoked, the system uses the particle number density balance equation and the gas-liquid mass balance equation as its core, and employs the fourth-order Runge-Kutta method to discretize and solve the partial differential equations. Vertically, the fog layer is divided into 20 layers, each 5 meters thick, to cover the typical height of radiation fog. The particle number density balance equation describes the spatiotemporal evolution of droplet number concentration, and the formula is: The first term on the right-hand side of the equation represents the decrease in number density caused by the collision-merger process. For collision-coupling kernel function, Long collision-coupling kernel form is adopted, N(r) is the number density distribution of droplets at the corresponding radius; the second term on the right is the number density flux divergence term caused by gravity settling, and vs is the settling velocity of droplets at the corresponding radius.
[0048] The gas-liquid mass balance equation describes the change in liquid water content, and the formula is: The first term on the right side of the formula represents the change in liquid water content caused by condensation growth (condensation growth term). The single-droplet condensation growth rate is derived from the water vapor diffusion equation; the second term on the right is the liquid-water mass redistribution term (collision-merging term) in the collision-merging process. The density of liquid water is taken as 1000 kg / m³. The collision-coalescence efficiency is a superposition of two mechanisms: gravitational settling and turbulent shearing. The total turbulent collision-coalescence efficiency is calculated as Et = Egravity + Eturbulence, where Egravity is calculated based on the difference in settling velocities of particles with different sizes, and Eturbulence is calculated based on empirical formulas for turbulent dissipation rate and particle inertial parameters.
[0049] After each time step, the droplet scale distribution is updated using the formula f(r,t+Δt)=f(r,t)+(condensation growth term+collision and coalescence term)*Δt. The liquid water content and droplet number density of the entire layer are obtained by integrating the updated scale distribution. Finally, these values are substituted into the visibility diagnostic equation to obtain the predicted horizontal visibility and fog dissipation rate. To ensure real-time performance, the Reuge mode solution utilizes GPU parallel acceleration, processing the spectral distribution calculations for each height layer in parallel to ensure that the entire fog field is calculated within a single 1-second time step.
[0050] In step S130, after obtaining the evaluation results, the system executes the target comparison and optimization initialization process. First, the predicted horizontal visibility value V(t+Δt) is compared with the preset target threshold Vtarget in the task parameters. If the predicted value has reached or exceeded the target threshold, the system determines that the current spraying parameters can meet the fog elimination requirements, and then switches to the maintenance monitoring state, reducing the sensor sampling frequency to once every 10 seconds, closing the optimization calculation link, and only continuously tracking the fog field changes. Once a downward trend in visibility is detected, the entire process is immediately reactivated.
[0051] If the predicted value fails to meet the target, the cumulative operation time tcum is further determined. If the cumulative time does not exceed the maximum allowable time Tmax, the system officially initiates the parameter optimization process. Upon startup, a dual-weighted objective function J, incorporating both fog removal efficiency and time consumption cost, is first constructed. The formula is as follows: In the formula, R(Ttarget) is the predicted fog removal rate that can be achieved within the target time, Ttarget is the predicted time required to achieve the visibility target, and α and β are weighting coefficients, which are 0.6 and 0.4 respectively, satisfying α+β=1. The first term [1-R] in this function represents the gap in fog removal effect. The higher the removal rate, the smaller the value of this term, guiding the algorithm to prioritize high fog removal efficiency. The second term [Ttarget / Tmax] represents the time cost. The shorter the time to reach the target, the smaller the value of this term, guiding the algorithm to choose a faster control path under similar efficiency. The value of α is higher than that of β, reflecting the control priority of prioritizing effect and taking into account timeliness.
[0052] After the objective function is constructed, the actual elevation angle θ(t) and azimuth angle of the current actuator will be calculated. The feeder speed ωf(t) and the fan speed ωb(t) are used as the initial values for optimization, and independent initial learning rates are configured for each of the four parameters. The learning rate for the gimbal elevation angle ηθ is set to 0.5° per iteration, and the learning rate for the azimuth angle is... The learning rate ηf for the feeder speed is set to 0.2° per iteration step, the learning rate ηb for the fan speed is set to 2% of the rated speed per iteration step, and the learning rate ηb for the fan speed is set to 2% of the rated speed per iteration step, to adapt to the different physical response sensitivities of each parameter.
[0053] Reference Figure 2 In S130, the steps of calling the multivariable gradient descent optimization algorithm to solve for the minimum value of the objective function and obtain the increment of the four-dimensional control parameters specifically include S210-S240: S210, initialize the parameter values of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed as the current optimization parameters; S220: Based on the current optimized parameters, use the microphysics evaluation model to predict visibility changes in future periods and generate a first visibility time series curve; in parallel, keep the optimized parameter values of the previous cycle unchanged, predict visibility changes in the corresponding future periods, and generate a second visibility time series curve. S230, calculate the difference between the first visibility time series curve and the second visibility time series curve. When the marginal improvement rate of the difference is lower than the preset efficiency threshold, it is determined that the efficiency boundary has been reached and the feeder speed increment is limited. S240 uses gradient descent to iteratively solve for the parameters that minimize the objective function, such as the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed, to obtain the updated current optimization parameters. These parameters are then used to generate the next round of visibility time series curves until the convergence condition is met.
[0054] Specifically, after the optimization process is initiated, the system synchronously executes counterfactual inference calculations. This step is implemented through dual-core parallel threads. The main thread uses the optimization parameters of the current iteration as the input boundary, calls the microphysics evaluation model to extrapolate the visibility evolution over the next 30 seconds, and generates the first visibility time series curve Vcurrent(t+Δt,t+2Δt,...,t+30Δt). The other independent parallel thread uses the complete parameter set after convergence in the previous cycle without applying any parameter adjustments. Based on the exact same initial fog field conditions, it extrapolates the visibility changes over the next 30 seconds and generates the second visibility time series curve Vprevious(t+Δt,t+2Δt,...,t+30Δt).
[0055] The 30-second simulation window is set to fully reflect the cumulative defogging effect of parameter adjustments and eliminate the interference of instantaneous fluctuations on efficiency judgment. Both time series use a 1-second calculation step size and share the same set of initial environmental parameters to ensure consistency of the comparison benchmark. Parallel threads use time-slicing scheduling to complete the simulation during the intervals of the main optimization calculation, without occupying the core computing resources of the main link, ensuring that the entire process is completed within a 1-second system cycle. The visibility difference at each time step is calculated using the formula ΔVCF(t+iΔt)=Vcurrent(t+iΔt)-Vprevious(t+iΔt), where i ranges from 1 to 30, and is used for subsequent marginal efficiency calculations.
[0056] After completing the simulation of two trajectories, the system performs efficiency boundary determination. First, it calculates the average difference between the two visibility time series within the entire simulation window. Then, it performs differential calculation on the average difference of multiple consecutive periods to obtain the marginal improvement rate. The calculation formula is d(ΔVCF) / dt=[ΔVCFmean(t)-ΔVCFmean(t-Δt)] / Δt, where ΔVCFmean is the average visibility difference within the simulation window, and Δt is the system calculation step size, i.e., 1 second.
[0057] The preset efficiency threshold δ is set to 0.01 m / s. When the marginal improvement rate is lower than this threshold, it is determined that the current dispensing parameters are close to the optimal efficiency boundary, and further increasing the catalyst dosage will have very little benefit in improving visibility. After the determination takes effect, the system immediately sends a feeder speed control signal to the gradient descent optimizer. First, the learning rate of the feeder speed dimension is reduced to 0.5 times the original value. At the same time, the maximum increment of the feeder speed in this iteration is limited to no more than 2% of the current value. If the condition that the marginal improvement rate is lower than the threshold is met for two consecutive calculation cycles, the target value of the feeder speed is further adjusted downward by 5%, actively reducing the catalyst supply and avoiding material waste caused by ineffective dispensing from the source of the algorithm. The entire determination and control logic is executed synchronously with the main optimization process without introducing additional computational delay.
[0058] The system then performs gradient descent iterative optimization, employing a sequential optimization strategy. It solves for the gimbal elevation angle, gimbal azimuth angle, fan speed, and feeder speed in a fixed order. After each parameter update, the current state is recalculated using the microphysical model, serving as the input for the next parameter optimization. The optimization process uses the finite difference method to calculate the partial derivatives of the objective function with respect to each parameter. This involves applying a small perturbation Δx to the parameter to be optimized, and obtaining the gradient value by the ratio of the change in the objective function ΔJ to the perturbation Δx. Then update the parameter values according to the gradient descent update rule.
[0059] The specific update formulas for the four parameters are as follows: gimbal elevation angle Gimbal azimuth feeder speed Fan speed The azimuth update rule additionally incorporates wind disturbance feedforward compensation. This forms a composite control structure of feedback optimization and feedforward disturbance rejection. The feedforward is responsible for quickly offsetting steady-state wind disturbances, while the feedback is responsible for finely correcting the remaining errors. The update of the feeder speed is constrained by the efficiency boundary of counterfactual deduction. When the boundary is reached, the step size is automatically reduced to avoid over-spreading.
[0060] After each complete update of the four-dimensional parameters, the system recalculates the objective function value and compares it with the previous round. When the decrease in the objective function is less than 0.001 after three consecutive iterations, the convergence condition is met and the iteration stops, outputting the final optimized parameters. If the iteration count exceeds 20 and convergence is still not achieved, the system is forcibly stopped and the current optimal solution is output, ensuring that the entire optimization process is completed within a 1-second calculation step, meeting the requirements of real-time closed-loop.
[0061] The steps of S240 in iteratively solving for the four-dimensional control parameters using gradient descent update rules also include: The target fog layer height is determined based on the fog top height in the fog field environmental parameters. Combined with the horizontal coverage range in the fog dissipation task parameters, the optimal combination of gimbal elevation angle and fan speed is solved by using the particle ballistic model and combining the particle size and drag coefficient in the catalyst physical property parameters. The lateral drift during particle flight is calculated based on the horizontal wind speed and wind direction deviation in the fog field environmental parameters, and a feedforward compensation is superimposed in the gimbal azimuth control loop to counteract the drift effect.
[0062] Specifically, while performing gradient descent optimization, the system simultaneously executes inverse kinematics calculations of the ballistic model, providing physical constraints for parameter optimization. The system first extracts the fog top height Htop from the fog field environmental parameters as the target fog layer height Htarget. Combining this with the target horizontal coverage range Ltarget from the task parameters, and using the median catalyst particle size dp as the reference particle, the system substitutes these values into the differential equations for the horizontal and vertical motion of the particles. These equations comprehensively consider air resistance, gravity, and buoyancy effects. The horizontal motion equation is as follows: The equation of vertical motion is In the formula, m is the mass of a single particle, vx and vz are the horizontal and vertical velocity components of the particle, respectively, Cd is the air resistance coefficient, which adopts the empirical formula of a standard sphere to adapt to different Reynolds number ranges, ρair is the air density, which is taken as 1.29 kg / m³ at normal temperature and pressure, and g is the acceleration due to gravity.
[0063] The terminal settling velocity of the particles is derived through force balance, and the formula is as follows: The horizontal range and height of the particle are obtained by integrating over the velocity; the formula for the horizontal range is: The formula for reaching the altitude is: In the formula, tland is the total flight time of the particle from its ejection to its settling at the target height or landing, and vb is the wind speed at the fan outlet, which is calculated from the fan speed.
[0064] The inverse kinematics process employs a binary iterative method. To improve real-time performance, the system pre-generates a three-dimensional lookup table for different particle sizes, target heights, and ranges. During runtime, an initial solution is obtained by interpolation through the lookup table, followed by 2 to 3 iterations for refinement. Within milliseconds, the optimal combination of elevation angle θ and fan speed ωb that satisfies the target height and coverage requirements is obtained. This result serves as a constraint benchmark for the corresponding parameters in gradient descent optimization, ensuring that the optimized output parameters always conform to the physical laws of particle trajectory and avoiding invalid solutions such as jet overshoot or insufficient range.
[0065] At the same time, the system performs wind disturbance feedforward compensation calculations, first extracting the horizontal wind speed U and wind direction deviation angle from environmental parameters. The wind direction deviation angle is defined as the angle between the current wind direction and the line connecting the center of the target fog area. Then, the flight time tflight of the particles from the device outlet to the center of the target fog area is calculated. This flight time is obtained by dividing the target horizontal distance Ltarget by the average of the particle's initial horizontal velocity. The initial horizontal velocity of the particles is calculated from the outlet wind speed vb corresponding to the current fan speed and the elevation angle θ. The lateral drift of particles is calculated based on the flight time and the crosswind speed component, using the following formula: The drift direction is consistent with the wind direction. Then, the required azimuth angle is calculated by using the arctangent relationship between the drift amount and the target distance. The formula is: The compensation direction is against the wind; that is, the azimuth angle is pre-deflected to the side from which the wind is blowing, ensuring that the particles, after being ejected, drift precisely to the center of the target fog area. This compensation is directly superimposed on the azimuth control command, responsible for offsetting the main component of steady-state wind disturbance. The gradient descent feedback adjustment is responsible for correcting the remaining pulsating wind disturbance and model error. Together, the two can control the landing point deviation caused by wind disturbance to within 5% of the characteristic scale of the fog area.
[0066] The steps of S240 in iteratively solving for the four-dimensional control parameters using gradient descent update rules also include: Set physical boundary constraints for the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed; The gradient descent iterative calculation results are forcibly restricted to the range of physical boundary constraints to ensure that the control instructions are within the executable range.
[0067] Specifically, to ensure the physical feasibility of all control commands, the system performs physical boundary constraints and pruning. The system sets hard constraint ranges for the four control parameters: the gimbal elevation angle constraint range is θ∈[45°, 85°], with a lower limit of 45° to ensure the jet has sufficient height to cover the upper part of the fog layer, avoiding catalyst settling and waste in the near region; and an upper limit of 85° to prevent near-vertical injection from causing particle aggregation near the device, while also preventing the jet from exceeding the fog top height; the gimbal azimuth angle constraint range is... It covers the entire circumferential working range and supports fog zone aiming in any direction; the feeder speed constraint range is... In the formula, Qmin is the minimum spreading mass rate to ensure effective dispersion, Qmax is the maximum conveying mass rate of the feeding mechanism, and kf is the speed-flow rate ratio coefficient. The lower limit prevents the catalyst from failing to disperse effectively due to excessively low spreading rates, resulting in material interruption or agglomeration. The upper limit prevents mechanical overload of the feeding mechanism and avoids secondary condensation caused by excessively high catalyst concentration per unit volume, which would reduce the demisting efficiency. The fan speed constraint range is ωb∈[30% rated speed, 100% rated speed]. The lower limit ensures that the particles obtain sufficient initial velocity to form an effective range, and the upper limit is the rated mechanical limit of the equipment to prevent overspeed from causing failure.
[0068] After updating the parameters in each iteration of gradient descent, the system immediately performs a boundary clipping operation. The clipping logic is xclipped=max(xmin,min(x,xmax)). Values exceeding the upper limit are forcibly set to the upper limit value, and values below the lower limit are forcibly set to the lower limit value. After clipping is completed, the next iteration is entered to ensure that all output control commands are within the physical range that the device can execute, and all boundary parameters can be flexibly configured through the industrial control interface according to the device model and catalyst type.
[0069] Furthermore, in S130, the step of comparing the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters also includes: When the predicted horizontal visibility is less than the preset target threshold and the cumulative operation time exceeds the maximum allowable time, the emergency response strategy is triggered. The feeder speed and fan speed are forcibly set to their respective upper limits to perform extreme fogging intervention.
[0070] Specifically, the system also sets up an emergency response fallback mechanism, which simultaneously verifies two conditions in the target determination stage of each cycle: first, whether the horizontal visibility prediction value is still lower than the target threshold; and second, whether the cumulative operation time has reached or exceeded the maximum allowable time. The cumulative operation time starts counting from the official start of seeding by the system and stops counting during the suspension of operation.
[0071] When both conditions are met simultaneously, it indicates that the conventional optimization strategy cannot achieve the fog elimination target within the allowed time. The system immediately triggers an emergency response strategy, suspending the normal gradient descent optimization process and directly forcing the feeder speed to the maximum upper limit value ωf_emergency=1.2*Qmax / kf, the fan speed to 100% of the rated speed, while keeping the gimbal elevation and azimuth angles unchanged at the current optimal aiming angles. Extreme fog elimination intervention is implemented with the maximum material delivery rate and maximum airflow energy. Simultaneously with triggering the emergency mode, a prominent emergency mode prompt pops up on the industrial control interface, informing the operator of the current status. The system reassesses the fog visibility every 30 seconds. Once the predicted visibility reaches the target value, it immediately exits the emergency mode and resumes conventional optimization control to avoid unnecessary catalyst waste and equipment damage caused by continuous extreme operations.
[0072] In S140, after the step of collecting and feeding back the measured data after execution to the microphysics evaluation model as the initial condition for the next iteration, the following steps are also included: Obtain the measured visibility sequence and liquid water content sequence; Determine the current call mode type: If the current mode is Kunkel, then the least squares fitting algorithm is used to dynamically correct the dispersion efficiency and effective diffusion volume parameters in the microphysical evaluation model based on the measured visibility sequence and liquid water content sequence. If the current mode is Reuge, the least squares fitting algorithm is used to dynamically correct the turbulence dissipation coefficient and collision efficiency factor in the microphysical evaluation model based on the measured visibility sequence and liquid water content sequence. The corrected parameters are then applied to subsequent calculations.
[0073] Specifically, in step S140, after parameter optimization and constraint verification are completed, the system enters the execution drive and data feedback stage. The four-dimensional control instruction set output by the decision control module is transmitted to the execution drive module through the real-time industrial Ethernet bus. The four independent PID controllers inside the module receive the set values of the corresponding parameters and synchronously drive the actuator to move.
[0074] The standard control law formula for a PID controller is: In the formula, e(t) is the deviation between the set value and the actual feedback value, Kp is the proportional coefficient, Ki is the integral coefficient, and Kd is the derivative coefficient. The PID parameters of each channel are pre-matched to the response characteristics of the corresponding actuator through a self-tuning algorithm.
[0075] The gimbal pitch servo channel receives the pitch angle setpoint and drives the pitch axis servo motor. Combined with position feedback from a 17-bit absolute encoder, it achieves a high-precision closed-loop control with a pointing accuracy better than ±0.5° and a response time within 100ms. The gimbal azimuth servo channel receives the azimuth angle setpoint after wind disturbance compensation. In addition to encoder feedback, an electronic magnetic compass is connected for absolute angle verification, eliminating accumulated errors from long-term operation. The feeder frequency conversion drive unit drives the screw feeder's frequency conversion motor, achieving a speed closed-loop control through a speed encoder, with a feeding rate control accuracy of ±3% of the setpoint. The fan frequency conversion drive unit adopts a cascade control structure, with the inner loop locking the motor speed and the outer loop correcting the speed setpoint based on feedback from the outlet wind speed sensor, achieving a wind speed control accuracy better than ±2% of the setpoint.
[0076] The four actuators receive commands and complete adjustments synchronously, ensuring the spatial and temporal coordination of the seeding parameters. After the execution is completed, sensors such as visibility meters and droplet spectrometers deployed in the work area continuously collect measured data, outputting a set of measured visibility sequences (Vobs) and measured liquid water content sequences (LWCobs) every second. These data are then transmitted back to the microphysics evaluation engine through the feedback calibration module as the initial conditions for the next round of iterative calculations, forming a complete closed-loop control chain.
[0077] After each round of closed-loop feedback, the system performs online calibration of the mode coefficients. The feedback calibration module continuously caches the measured visibility and liquid water content time series. When the length of the cached continuous valid data reaches 30 seconds, and the data shows no abnormal jumps and the fluctuations are stable, the coefficient calibration process is automatically started in a low-priority background thread, without occupying the computing resources of the main control link. Before calibration, the current evaluation mode is first determined. If the current mode is Kunkel, the dispersion efficiency correction coefficient and the effective diffusion volume correction coefficient are selected as the parameters to be calibrated; if the current mode is Reuge, the turbulence dissipation coefficient is selected. Collision efficiency correction factor As parameters to be calibrated. The calibration aims to minimize the sum of squared residuals between the model's calculated and measured values over the past 30 seconds, constructing a least-squares fitting optimization problem with the objective function as follows: In the formula, Vmode is the visibility value calculated by the model, Vobs is the measured visibility value, and N is the number of calibration sample points.
[0078] The Levenberg-Marquardt algorithm is used to solve the problem. The parameters to be calibrated are iteratively adjusted to continuously reduce the sum of squared residuals. The convergence criterion is that the decrease in residuals between two consecutive iterations is less than 1e-6. After calibration, the system compares the new coefficients with the original coefficients. If the deviation is within 5%, the calibration is considered valid, and the new coefficients are added to the algorithm parameter library. If the deviation is too large, the data is marked as suspicious, the original coefficients are retained, and the running log is recorded for subsequent manual verification, thus preventing abnormal data from degrading model accuracy.
[0079] Finally, the calibration parameters are applied. The effective coefficients obtained from the calibration are immediately updated in the parameter register of the evaluation engine and automatically loaded and taken effect in the next 1-second iteration cycle, replacing the original default coefficients or the previous version of the calibration coefficients. At the same time, the system will classify and store the calibration coefficients according to the fog field type and application scenario, such as distinguishing between fog types such as radiation fog and advection fog, and scenarios such as airports, ports, and highways, forming a scenario-based parameter library. When the same type of fog field and scenario is detected again, the system will automatically load the corresponding historical calibration coefficients as initial values, which will significantly reduce the convergence time of on-site calibration.
[0080] The parameter updates employ an incremental mechanism, adjusting only the corresponding coefficients for each mode during each calibration, without affecting other modes or parameters. Furthermore, the system permanently retains coefficient values from at least three historical versions, supporting one-click rollback to previous versions to ensure system stability and reliability. Through this continuous online calibration and iterative updates, the prediction accuracy of the microphysics model gradually improves with the number of calibrations, continuously reducing the systematic deviation between numerical simulations and actual fog fields, achieving adaptive evolution and continuous optimization of the model.
[0081] Based on the above method embodiments, the second embodiment of this application discloses an automatic warm fog dispersing and spraying control system based on a microphysical evaluation model. The automatic warm fog dispersing and spraying control system based on a microphysical evaluation model of this application embodiment can implement any of the above-mentioned automatic warm fog dispersing and spraying control methods based on microphysical evaluation models, and the specific working process of each module in the automatic warm fog dispersing and spraying control system based on a microphysical evaluation model can be referred to the corresponding process in the above method embodiments.
[0082] Reference Figure 3 For ease of understanding, an example is provided below: An automatic warm fog dispersal control system based on a microphysical evaluation model includes: The data acquisition module is used to receive fog removal task parameters, acquire fog field environmental parameters, catalyst physical property parameters and actuator status parameters of the fog removal device in real time, and verify the validity of the parameters. The microphysics assessment engine module is used to determine the fog field type based on the fog field environmental parameters after verification, adaptively call the Kunkel mode or Reuge mode to build a microphysics assessment model, calculate the evolution parameters of fog field microphysics process in real time, and output the predicted values of horizontal visibility and fog layer elimination rate at future times based on the evolution parameters of fog field microphysics process. The decision control module is used to compare the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, it constructs an objective function containing fog dissipation efficiency and time consumption cost based on the microphysical evaluation model, and calls the multivariate gradient descent optimization algorithm to solve for the minimum value of the objective function, thereby obtaining the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. The execution drive module is used to generate control commands based on the incremental four-dimensional control parameters to drive the fog dispersing device to perform the spreading action. The feedback calibration module is used to collect the measured data after execution and feed it back to the microphysics evaluation engine module as the initial conditions for the next iteration.
[0083] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce a good effect.
[0084] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.
Claims
1. A method for automatic spraying control of warm fog dispersal based on a microphysical evaluation model, characterized in that, include: Receive fog removal task parameters, and acquire fog field environmental parameters, catalyst physical property parameters, and actuator status parameters of the fog removal device in real time, and verify the validity of the parameters; Based on the verified fog field environmental parameters, the fog field type is determined, and a microphysical evaluation model is established by adaptively calling the Kunkel mode or Reuge mode. The evolution parameters of the fog field microphysical process are calculated in real time, and based on the evolution parameters of the fog field microphysical process, the predicted values of horizontal visibility and fog layer elimination rate at future times are output. The predicted horizontal visibility value is compared with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, an objective function containing fog dissipation efficiency and time consumption cost is constructed based on the microphysical evaluation model, and the multivariate gradient descent optimization algorithm is called to solve for the minimum value of the objective function, so as to obtain the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. Based on the incremental generation of the four-dimensional control parameters, control commands are generated to drive the fog dispersing device to perform the spreading action, and the measured data after execution is collected and fed back to the microphysical evaluation model as the initial conditions for the next iteration.
2. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 1, characterized in that, The steps involved in determining the fog field type based on the verified fog field environmental parameters, adaptively calling Kunkel or Reuge mode to establish a microphysical evaluation model, and calculating the evolution parameters of the fog field's microphysical processes in real time include: Based on the aforementioned fog field environmental parameters, three indices—turbulent dissipation rate, vertical wind speed root mean square error, and fog droplet spectrum fitting goodness of fit—are jointly determined for evaluation. When the turbulent dissipation rate is less than the first threshold, the vertical wind speed root mean square error is less than the second threshold, and the goodness of fit of the fog droplet spectrum is greater than the third threshold, the fog field is determined to be a homogeneous weak turbulent state, and the Kunkel model is called to calculate the water vapor consumption. If any of the criteria is not met, the fog field is determined to be in a heterogeneous turbulent state, and the process is switched to Reuge mode to solve the particle number density balance equation and the gas-liquid mass balance equation.
3. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 2, characterized in that, The steps for outputting predicted horizontal visibility and fog dissipation rates for future times based on the aforementioned fog field microphysical process evolution parameters specifically include: When Kunkel mode is invoked, the feed mass flow rate, dispersion efficiency, and effective diffusion volume are determined based on the actuator state parameters. Combined with the particle mass in the catalyst physical property parameters, the catalyst number density per unit volume is calculated. The mass change during particle descent is calculated based on the water vapor diffusion equation. The water vapor consumption per unit volume is obtained by combining the catalyst number density and the mass change. The water vapor consumption per unit volume is then substituted into the visibility diagnostic equation to output the horizontal visibility prediction value and the fog elimination rate prediction value. or, When the Reuge mode is invoked, the particle number density balance equation and the gas-liquid mass balance equation, which include the turbulence dissipation coefficient and the collision efficiency factor, are solved to obtain the updated droplet size distribution and liquid water content. The updated liquid water content is then substituted into the visibility diagnostic equation to output the predicted horizontal visibility and the predicted fog layer elimination rate.
4. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 1, characterized in that, The specific steps for using the multivariable gradient descent optimization algorithm to solve for the minimum of the objective function and obtain the four-dimensional control parameter increments include: Initialize the parameters of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed as the current optimization parameters; Based on the current optimization parameters, the visibility changes in future time periods are predicted using the microphysical evaluation model, and a first visibility time series curve is generated. In parallel, while keeping the parameter values optimized in the previous cycle unchanged, the visibility changes in the corresponding future time period are predicted to generate a second visibility time series curve. Calculate the difference between the first visibility time series curve and the second visibility time series curve. When the marginal improvement rate of the difference is lower than a preset efficiency threshold, it is determined that the efficiency boundary has been reached and the feeder speed increment is limited. The gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed parameters that minimize the objective function are obtained by iteratively solving the rule update through gradient descent. The updated current optimization parameters are then used to generate the next round of visibility time series curves until the convergence condition is met.
5. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 4, characterized in that, The steps for iteratively solving the four-dimensional control parameters using gradient descent update rules also include: The target fog layer height is determined based on the fog top height in the fog field environmental parameters. Combined with the horizontal coverage range in the fog dissipation task parameters, the optimal combination of gimbal elevation angle and fan speed is solved by using a particle ballistic model and combining the particle size and drag coefficient in the catalyst physical property parameters. The lateral drift during particle flight is calculated based on the horizontal wind speed and wind direction deviation in the fog field environmental parameters, and a feedforward compensation is superimposed in the gimbal azimuth control loop to counteract the drift effect.
6. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 5, characterized in that, The steps for iteratively solving the four-dimensional control parameters using gradient descent update rules also include: Set physical boundary constraints for the gimbal elevation angle, gimbal azimuth angle, feeder speed, and fan speed; The gradient descent iterative calculation results are forcibly restricted to the range of physical boundary constraints to ensure that the control instructions are within the executable range.
7. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 3, characterized in that, After the step of collecting and feeding back the measured data after execution to the microphysics evaluation model as the initial condition for the next iteration, the following steps are also included: Obtain the measured visibility sequence and liquid water content sequence; Determine the current call mode type: If the current mode is Kunkel, then the least squares fitting algorithm is used to dynamically correct the dispersion efficiency and effective diffusion volume parameters in the microphysical evaluation model based on the measured visibility sequence and the liquid water content sequence. If the current mode is Reuge, then the least squares fitting algorithm is used to dynamically correct the turbulence dissipation coefficient and collision efficiency factor in the microphysical evaluation model based on the measured visibility sequence and the liquid water content sequence. The corrected parameters are then applied to subsequent calculations.
8. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 1, characterized in that, The step of comparing the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters further includes: When the predicted horizontal visibility value is less than the preset target threshold and the cumulative operation time exceeds the maximum allowable time, an emergency response strategy is triggered. The feeder speed and fan speed are forcibly set to their respective upper limits to perform extreme fogging intervention.
9. The automatic spraying control method for warm fog dispersal based on a microphysical evaluation model according to claim 1, characterized in that, The specific steps for validating parameters include: The physical reasonable threshold range of the fog field environmental parameters, the catalyst physical property parameters, and the actuator status parameters is verified. When parameters exceed limits, an alarm is triggered and the system is prevented from starting. After verification, the ambient temperature, air pressure, liquid water content, and wind field parameters are standardized and packaged as input boundary conditions for the microphysical evaluation model.
10. An automatic spraying control system for warm fog dispersal based on a microphysical evaluation model, characterized in that, The method for automatic warm fog dispersal control based on a microphysical evaluation model as described in any one of claims 1 to 9 includes: The data acquisition module is used to receive fog removal task parameters, acquire fog field environmental parameters, catalyst physical property parameters and actuator status parameters of the fog removal device in real time, and verify the validity of the parameters. The microphysics assessment engine module is used to determine the fog field type based on the fog field environmental parameters after verification, adaptively call the Kunkel mode or Reuge mode to build a microphysics assessment model, calculate the evolution parameters of the fog field microphysics process in real time, and output the predicted values of horizontal visibility and fog layer elimination rate at future times based on the fog field microphysics process evolution parameters. The decision control module is used to compare the predicted horizontal visibility value with the preset target threshold in the fog dissipation task parameters. When the predicted value does not meet the target, it constructs an objective function containing fog dissipation efficiency and time consumption cost based on the microphysical evaluation model, and calls the multivariate gradient descent optimization algorithm to solve for the minimum value of the objective function, thereby obtaining the four-dimensional control parameter increments of gimbal elevation angle, gimbal azimuth angle, feeder speed and fan speed. The execution drive module is used to generate control commands based on the incremental four-dimensional control parameters to drive the defogging device to perform the spreading action. The feedback calibration module is used to collect the measured data after execution and feed it back to the microphysics evaluation engine module as the initial condition for the next iteration.
