A mobile intelligent control spray dust suppression system

Abstract

The present invention discloses a mobile intelligent-controlled spray dust suppression system, specifically relating to the technical field of compensating for or utilizing the external environment when a spray dust suppression vehicle is in wind field disturbance conditions. This system involves utilizing multiple wind speed and direction sensing nodes mounted on a lifting mast to collect local wind field parameters at multiple different heights and directions while the spray dust suppression vehicle is in motion, constructing a wind speed gradient dataset based on the local wind field parameters, and performing normalization and disturbance feature extraction on the wind speed gradient dataset to generate a wind disturbance time series and spatial disturbance distribution map. By constructing a time-space coupled control mechanism that integrates wind disturbance feature extraction, trend prediction, and path calibration, this system achieves early perception of non-steady-state wind field disturbances and dynamic correction of spray response parameters, thereby resolving the issues of perception-control coupling lag and spray offset error.

Description

Technical Field

[0001] The present invention relates to the technical field of spray dust suppression vehicles compensating for or utilizing the external environment under wind field disturbance conditions, and more specifically, to a mobile intelligently controlled spray dust suppression system. Background Art

[0002] Mobile spray dust suppression systems, i.e., spray-type dust suppression vehicles, operating in environments with frequently changing wind speeds face the core problem of coupling lag between traditional perception and control.

[0003] In the existing technology, spray dust suppression vehicles generally adopt a feedback control mechanism based on real-time wind speed monitoring and wind direction analysis. That is, meteorological sensors sense the current wind field data in real time and adjust the nozzle angle, spray pressure or spray volume accordingly.

[0004] However, in areas with frequent microclimate disturbances and prone to sudden changes in wind speed and direction, such as coastal ports, canyon wind outlets, and high-rise construction zones, this type of control strategy exhibits significant limitations. Specifically, the wind field structure is highly non-stationary, with air disturbances exhibiting discontinuous spatial distribution and temporal transitions. The spray dust suppression system of the spray truck, based on its "current perception-inferred control" strategy, suffers from a persistent response delay, making it difficult for the spray trajectory to align with the dynamic center of the dust accumulation area.

[0005] Furthermore, most current systems treat the wind field sensing and spray execution chain as a linear causal structure, ignoring the nonlinear feedback effect of the wind field itself on the fog cloud propagation path. This "perception lag-control sluggishness" mechanism creates a significant fog cloud escape effect, especially in the initial spraying phase: the spray is carried away from the target area by rapidly changing wind disturbances, resulting in persistent offset errors and spraying dead zones, severely reducing dust reduction efficiency and resource utilization, and even causing secondary dust suspension in areas with long slopes or high towers.

[0006] Therefore, the core problem faced by the system is not simply the untimely adjustment of wind speed, but the lack of active prediction capabilities for future disturbance trends in the wind field and the structural pre-control mechanism of air propagation paths. This constitutes an obstacle for the spray system to operate in a non-steady-state environment. Its essence is a structural coupling error escape problem between perception and control. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a mobile intelligent control spray dust suppression system, which realizes early perception of non-steady-state wind field disturbances and dynamic correction of spray response parameters by constructing a time-space coupling control mechanism that integrates wind disturbance feature extraction, trend prediction and path calibration, thereby solving the perception-control coupling lag and spray offset error problems existing in the background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solutions: a mobile intelligent control spray dust suppression system, comprising an intelligent control system, an environmental treatment system, and a spraying vehicle, wherein the intelligent control system and the environmental treatment system are installed on the spraying vehicle, and the environmental treatment system comprises an atomizing nozzle, a water storage tank, a power supply system, and a lifting rod; the intelligent control system comprises a wind field construction module, a feature extraction module, a prediction module, a calibration module, and a compensation module;

[0009] The wind field construction module is used to collect local wind field parameters at different heights and directions using multiple wind speed and direction sensor nodes installed on the lifting mast while the spray dust suppression sprayer is moving, and to construct a wind speed gradient dataset based on the local wind field parameters;

[0010] The feature extraction module is used to perform normalization and disturbance feature extraction processing on the wind speed gradient data set to generate a wind disturbance time series and a spatial disturbance distribution map;

[0011] The prediction module is used to input the wind disturbance time series and spatial disturbance distribution map into the preset time-space dual memory nested transformer prediction model, and output the wind disturbance change trend prediction results and spray trajectory offset parameters in the target time domain;

[0012] The calibration module is used to control some atomizing nozzles to release droplets for testing wind disturbance. It also monitors the drift path of the droplets based on images or acoustic wave feedback devices, compares the drift path with the spray trajectory offset parameters, and updates the spray trajectory offset parameters.

[0013] The compensation module is used to generate control instructions for the posture adjustment, spray angle correction and spray pressure change of the atomizing nozzle according to the corrected spray trajectory offset parameters, and drive the environmental management system through the intelligent control system to complete the spray compensation control.

[0014] In a preferred embodiment, the intelligent control system also includes a closed-loop control module, which is used to re-collect the local wind field parameters of each wind speed and direction sensor node on the lifting rod after the compensation module completes the spray compensation control, and construct a new wind speed gradient data set; input the new wind speed gradient data set into the feature extraction module, and pass through the feature extraction module, prediction module, calibration module and compensation module in sequence; through the periodic linkage call of the intelligent control system, a closed-loop control mechanism of wind disturbance prediction and spray control is formed with changes in local wind field parameters as input and control instruction updates as output.

[0015] In a preferred embodiment, the atomizing nozzle is installed on the atomizing bracket at the rear of the spraying vehicle, and the atomizing nozzle is connected to the water storage tank through a liquid pipe. The power supply system is set at the chassis of the spraying vehicle and provides power output to the atomizing nozzle, lifting rod and related control circuits through a line distribution interface; the lifting rod is fixed to the bracket structure on the top of the spraying vehicle, and a wind speed and direction sensor node is integrated in the lifting rod. At the same time, the wind speed and direction sensor node is used as a front-end acquisition device for local wind field parameters, and the output data of the wind speed and direction sensor node is transmitted to the intelligent control system;

[0016] The intelligent control system and the environmental management system establish a two-way connection through the control harness and the data communication module. The intelligent control system is used to receive the local wind field parameters of the lifting rod and send control instructions for spraying to the atomizing nozzle according to the prediction module, calibration module and compensation module.

[0017] In a preferred embodiment, when the feature extraction module performs normalization and disturbance feature extraction processing on the wind speed gradient data set, the normalization processing includes performing mean shift and range scaling processing on the collected wind speed and wind direction parameters at each altitude and azimuth according to a preset time window, that is, forming a wind speed gradient matrix under a unified scale; after completing the normalization processing, the feature extraction module calculates the wind speed vector difference and wind direction rotation offset between adjacent measuring points at the same time section based on the wind speed gradient matrix, and counts the wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate within a specified time step as wind disturbance features to form a wind disturbance feature vector group;

[0018] The wind disturbance feature vector groups are arranged in chronological order to construct a wind disturbance time series. At the same time, the feature extraction module, based on the spatial position of each measuring point of the wind speed and direction sensor nodes, performs spatial interpolation on the wind disturbance features within the corresponding time window and maps them to the wind field area distribution map, thereby solving the local wind disturbance intensity gradient field and rotation direction distribution map to form a spatial disturbance distribution map.

[0019] In a preferred embodiment, the wind speed fluctuation rate of the wind disturbance characteristic is calculated by calculating the ratio of the standard deviation to the mean of the wind speed values ​​at the same measuring point within a specified time window to reflect the relative change amplitude of the wind speed within the time window;

[0020] The wind direction rotation frequency of the wind disturbance feature is measured by counting the number of times the wind direction changes exceed a preset angle threshold within a specified time window and normalizing it to the change frequency per unit time to measure the wind direction instability;

[0021] The wind speed derivative change rate of the wind disturbance feature is calculated by calculating the difference value of the wind speed derivative in adjacent time steps and performing statistics on the derivative change in unit time to represent the sudden change intensity of the wind speed change trend.

[0022] In a preferred embodiment, in the prediction module, based on the wind disturbance time series and spatial disturbance distribution map generated by the feature extraction module, the wind speed fluctuation rate, wind direction rotation frequency, and wind speed derivative change rate are used to form a time input tensor at continuous time steps, and the wind disturbance intensity gradient field and rotation direction distribution map are used to form a spatial input tensor for constructing a joint input structure of the converter prediction model.

[0023] The time input tensor and the space input tensor are respectively input into the time memory encoding module and the space memory encoding module of the converter prediction model. The time memory encoding module extracts the temporal trend characteristics of wind disturbance changes based on the multi-head attention mechanism and the position nested structure. The spatial memory encoding module extracts the spatial correlation characteristics of wind disturbances based on the position mapping and the disturbance distribution weighting mechanism.

[0024] Performing feature fusion and residual connection operations on the temporal trend features and the spatial correlation features in a fusion control unit of a transformer prediction model to form a unified temporal-spatial coupling feature tensor;

[0025] The time-space coupling feature tensor is input into the decoding module of the converter prediction model. According to the preset target time domain sliding window, the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend in the future target time period are predicted, and the wind disturbance change trend prediction result is output;

[0026] Based on the wind disturbance change trend prediction results and the historical disturbance trajectory distribution in the spatial disturbance distribution map, the offset angle, offset direction and offset amplitude of the spray trajectory in the target time domain are calculated, and the spray trajectory offset parameters are output.

[0027] In a preferred embodiment, the decoding module of the transformer prediction model includes a position encoding unit, a multi-head attention mechanism layer, a feedforward neural network layer and an output mapping layer. The decoding module receives the time-space coupling feature tensor output by the fusion control unit, and combines the time index position of the current observed time step, and encodes the position index of each predicted time step in the target time period through the position encoding unit to generate a corresponding time position code, and embeds the generated time position code into the time-space coupling feature tensor;

[0028] The multi-head attention mechanism layer in the decoding module is used to construct an attention mapping relationship between historical features and predicted positions at the target time position, extract contextual feature representations between historical time steps corresponding to each predicted time step with weight values ​​within a preset upper threshold range, and use the feedforward neural network layer to perform nonlinear transformations on the contextual feature representations to output prediction intermediate results with time-recursive characteristics.

[0029] The output mapping layer of the decoding module projects the intermediate prediction results into the prediction component spaces of wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate, forming the wind disturbance change trend prediction results;

[0030] Among them, the preset target time domain sliding window slides backward from the current moment by a preset time length, and the preset time length corresponds to the number of future time steps set in the converter prediction model; the decoding module performs attention weight allocation operations on the fused time-space coupling feature tensor in sequence at each prediction time step within the target time domain sliding window, and inputs the attention weighted results into the feedforward neural network layer for feature mapping processing to generate the disturbance prediction output corresponding to the time step, and gradually generates a sequence of predicted values ​​of each disturbance variable within the target time period, thereby obtaining the output of the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend.

[0031] In a preferred embodiment, the atomizing nozzles in a local area where the wind speed fluctuation rate exceeds a preset wind speed fluctuation threshold, or the atomizing nozzles in a local area where the wind speed derivative change rate exceeds a preset wind speed derivative change threshold are defined as partial atomizing nozzles;

[0032] During the non-dust suppression operation phase, test droplets are released based on the preset disturbance intensity sorting rules. The spray of droplets adopts an intermittent spraying method with a duration of less than 1 second and a spray pressure lower than 50% of the lower limit of the normal operating pressure;

[0033] The image or sound wave return device includes an image recognition device or a sound wave positioning device, which is installed on a lifting rod to collect the movement process of the droplets and obtain the drift path of the droplets based on continuous frame matching and position inversion methods; the drift path is subjected to path overlap analysis with the spray trajectory offset parameter output by the prediction module, and the offset error between the drift path and the predicted trajectory is constructed to characterize the relative error between the predicted result and the actual disturbance response;

[0034] Based on the offset error, error correction is performed on the offset angle, offset direction and offset amplitude in the spray trajectory offset parameters in turn. The error correction includes spatially decomposing the offset error between the drift path and the predicted trajectory into corresponding offset direction angle and offset amplitude components, and adjusting the corresponding parameters respectively; at the same time, based on the time difference between the response moment of the droplet drift path and the timestamp corresponding to the predicted trajectory, the time offset is calculated, and the time offset is used as a reference for evaluating the prediction hysteresis.

[0035] The technical effects and advantages of the present invention are as follows:

[0036] This system introduces a wind disturbance trend prediction mechanism based on a time-space dual memory structure, actively models key disturbance factors such as wind speed fluctuation rate, wind direction rotation frequency, and wind speed derivative change rate. This enables early prediction of disturbance trends in unsteady wind fields, solves the core problem of coupled lag between perception and control in traditional feedback spray systems, and improves the ability of the spray trajectory to adapt to the dynamic distribution of dust.

[0037] The system introduces wind disturbance intensity gradient fields and rotation direction distribution maps into the wind field modeling process, achieving quantitative modeling of disturbance spatial distribution trends and dominant rotation directions, providing structured spatiotemporal disturbance inputs for the prediction model, thereby enhancing the model's ability to depict the evolution of wind disturbances in complex microclimate regions.

[0038] By using a short-term, low-intensity, micro-droplet release strategy, combined with image or acoustic wave return path reconstruction technology, the system can achieve dynamic sampling of the disturbance feedback path and actual response monitoring. Based on this, it constructs the spatial offset error between the predicted trajectory and the actual drift path, providing a reference for closed-loop correction of spray control parameters.

[0039] The wind disturbance intensity ranking rule is used to set the response priority of high-pressure atomizing nozzles, and local area spray disturbance tests are conducted to ensure more rational allocation of nozzle resources, avoid energy waste and droplet diffusion interference;

[0040] By building a control structure that integrates feature extraction, trend prediction, path verification and command compensation, the system can complete the prediction model update and control command reconstruction within the next control cycle after the wind field changes, ensuring that the spray control strategy can adapt to changes in environmental disturbances and maintain continuous and effective response. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the system module of the present invention.

[0042] Figure 2 This is a physical picture of the intelligent control system and environmental management system in the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] Refer to the instruction manual Figure 1-2A mobile intelligent control spray dust suppression system according to one embodiment of the present invention includes an intelligent control system, an environmental treatment system, and a spraying vehicle. The intelligent control system and the environmental treatment system are installed on the spraying vehicle. The spraying vehicle is a carrier for providing mobility. The spraying vehicle provides a basic transportation and energy docking platform for the intelligent control system and the environmental treatment system. The environmental treatment system includes a high-pressure atomizing nozzle, a water storage tank, a power supply system, and an automatic lifting rod. The intelligent control system includes a wind field construction module, a feature extraction module, a prediction module, a calibration module, and a compensation module.

[0045] The wind field construction module is used to collect local wind field parameters at different heights and directions using multiple wind speed and direction sensor nodes installed on the automatic lifting mast while the spray dust suppression sprayer is moving, and to construct a wind speed gradient dataset based on the local wind field parameters;

[0046] The feature extraction module is used to perform normalization and disturbance feature extraction processing on the wind speed gradient data set to generate a wind disturbance time series and a spatial disturbance distribution map;

[0047] The prediction module is used to input the wind disturbance time series and spatial disturbance distribution map into the preset time-space dual memory nested transformer prediction model, and output the wind disturbance change trend prediction results and spray trajectory offset parameters in the target time domain;

[0048] The calibration module is used to control some high-pressure atomizing nozzles to release trace droplets for testing wind disturbances. This is used for wind disturbance path testing and prediction result verification. The module also monitors the drift path of trace droplets based on image or sound wave return devices, compares the drift path with the spray trajectory offset parameters, and updates the spray trajectory offset parameters.

[0049] The compensation module is used to generate control instructions for posture adjustment, spray angle correction and spray pressure change of the high-pressure atomizing nozzle based on the corrected spray trajectory offset parameters, and drive the environmental management system through the intelligent control system to complete the spray compensation control.

[0050] The intelligent control system also includes a closed-loop control module, which is used to re-collect the local wind field parameters of each wind speed and direction sensor node on the automatic lifting rod after the compensation module completes the spray compensation control, and construct a new wind speed gradient data set; input the new wind speed gradient data set into the feature extraction module, and sequentially undergo normalization and disturbance feature extraction by the feature extraction module, trend prediction by the prediction module, update of spray trajectory offset parameters by the calibration module, and spray compensation control generation by the compensation module; through the periodic linkage call of the intelligent control system, a closed-loop control mechanism of wind disturbance prediction and spray control is formed with changes in local wind field parameters as input and control instruction updates as output.

[0051] The high-pressure atomizing nozzle is installed on the atomizing bracket at the rear of the spraying vehicle. The high-pressure atomizing nozzle is connected to the water storage tank through a liquid pipe. The power supply system is set at the chassis of the spraying vehicle and provides power output to the high-pressure atomizing nozzle, the automatic lifting rod and the related control circuit through the line distribution interface. The automatic lifting rod is fixed to the bracket structure on the top of the spraying vehicle. The automatic lifting rod is integrated with a wind speed and direction sensor node. At the same time, the wind speed and direction sensor node is used as a front-end acquisition device for local wind field parameters. The output data of the wind speed and direction sensor node is transmitted to the intelligent control system.

[0052] The intelligent control system and the environmental management system establish a two-way connection through the control harness and the data communication module. The intelligent control system is used to receive the local wind field parameters of the automatic lifting mast and send control instructions for spraying to the high-pressure atomizing nozzle according to the prediction module, calibration module and compensation module. The data communication module includes but is not limited to a main control communication interface, a data conversion and distribution unit, a wireless communication module and a signal synchronization and cache module, which is used to realize two-way transmission of data between the intelligent control system and the environmental management system.

[0053] The power supply of the intelligent control system and the environmental treatment system is connected to the main energy system provided by the spraying truck. The main energy system of the spraying truck includes a fuel power system or an electric power system (that is, a gasoline truck or an electric truck). The fuel power system of the spraying truck works by driving the integrated on-board power generation unit of the engine. The on-board power generation unit outputs regulated DC power, which provides stable power to the intelligent control system and the environmental treatment system through the power supply system; the electric power system is based on the power battery pack configured in the spraying truck, and distributes the required working voltage to the power supply system through the on-board power management module to realize direct energy supply to the intelligent control system and the environmental treatment system; the above two types of power systems correspond to fuel-type spraying trucks and electric-type spraying trucks respectively, and realize the connection of energy support and control coordination through a unified power supply system interface and control bus;

[0054] The high-pressure atomizing nozzle also integrates a liquid extraction execution component, a posture adjustment mechanism, a spray angle adjustment device and a spray pressure control unit, wherein the liquid extraction execution component is connected to the water storage tank through a liquid pipe, and is used to transport the liquid in the water storage tank to the nozzle of the high-pressure atomizing nozzle; the posture adjustment mechanism is connected to the bracket of the high-pressure atomizing nozzle through a mechanical connecting rod or an electric drive unit to realize the pitch adjustment function of the nozzle around the fixed axis; the spray angle adjustment device is arranged on the movable guide component provided at the front end of the nozzle, and is used to adjust the deflection angle of the spray direction under the action of the control signal. The movable guide component includes but is not limited to an electric servo swing mechanism, a dual-axis servo assembly, a stepper motor rotary joint or an electric universal ball seat assembly, etc.; the spray pressure control unit is connected to the power supply system and receives the pressure control instruction from the intelligent control system through the control harness, and is used to adjust the working pressure of the liquid atomization at the nozzle, thereby realizing spray compensation control under multi-dimensional linkage;

[0055] In addition, in actual applications, the automatic lifting rod can be an electric push rod lifting rod (linear electric cylinder), a synchronous belt driven lifting rod or a pneumatic lifting rod;

[0056] The wind speed and direction sensing nodes include but are not limited to integrated ultrasonic wind speed and direction sensors such as WindSonic, Vaisala-WXT536 or RM-Young series, which are suitable for high-frequency wind field monitoring of mobile spray platforms.

[0057] When the feature extraction module performs normalization and disturbance feature extraction processing on the wind speed gradient data set, the normalization processing includes performing mean shift and range scaling processing on the collected wind speed and wind direction parameters at each altitude and azimuth according to a preset time window, that is, forming a wind speed gradient matrix under a unified scale; after completing the normalization processing, the feature extraction module calculates the wind speed vector difference and wind direction rotation offset between adjacent measuring points at the same time section based on the wind speed gradient matrix, and counts the wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate within the specified time step as wind disturbance features to form a wind disturbance feature vector group;

[0058] The wind disturbance feature vector groups are arranged in chronological order to construct a wind disturbance time series. At the same time, the feature extraction module uses the spatial position of each measuring point of the wind speed and direction sensor nodes as a basis to spatially interpolate the wind disturbance features within the corresponding time window and map them to the wind field regional distribution map. The local wind disturbance intensity gradient field and rotation direction distribution map are solved to form a spatial disturbance distribution map, thus achieving a structured conversion from the wind speed gradient dataset to the spatiotemporal disturbance expression form, providing an input basis for the subsequent prediction module modeling.

[0059] The local wind disturbance intensity gradient field is based on the completion of the spatial interpolation of the wind speed fluctuation rate. The wind speed derivative change rate after interpolation is combined to calculate the spatial first-order partial derivative of the wind speed fluctuation rate and the local fluctuation intensity of the wind speed derivative change rate at each interpolation grid point. A composite gradient field representing the change rate of disturbance strength and mutation sensitivity is formed through vector superposition, which is used to describe the spatial distribution trend of wind disturbance intensity and the dynamic disturbance evolution characteristics in the local area.

[0060] The rotation direction distribution diagram is based on the interpolation results of the wind direction rotation frequency in the spatial grid, constructs the rotation frequency change vector between adjacent interpolation points, introduces the synchronization weight of the wind speed derivative change rate, and calculates the time stability index and spatial consistency index of the main rotation direction, which is used to comprehensively describe the spatial dominant direction of wind direction changes and the rotation trend evolution pattern in the disturbance area.

[0061] The wind speed fluctuation rate of the wind disturbance characteristic is calculated by calculating the ratio of the standard deviation to the mean of the wind speed values ​​at the same measuring point within a specified time window to reflect the relative change amplitude of the wind speed within the time window;

[0062] The wind direction rotation frequency of the wind disturbance feature is measured by counting the number of times the wind direction changes exceed a preset angle threshold within a specified time window and normalizing it to the change frequency per unit time to measure the wind direction instability;

[0063] The wind speed derivative change rate of the wind disturbance feature is calculated by calculating the difference value of the wind speed derivative in adjacent time steps and performing statistics on the derivative change in unit time to represent the sudden change intensity of the wind speed change trend.

[0064] In the prediction module, based on the wind disturbance time series and spatial disturbance distribution map generated by the feature extraction module, the wind speed fluctuation rate, wind direction rotation frequency, and wind speed derivative change rate are combined into a time input tensor at continuous time steps. At the same time, the wind disturbance intensity gradient field and rotation direction distribution map are combined into a spatial input tensor to construct a joint input structure for the converter prediction model.

[0065] The time input tensor and the space input tensor are respectively input into the time memory encoding module and the space memory encoding module of the converter prediction model. The time memory encoding module extracts the temporal trend characteristics of wind disturbance changes based on the multi-head attention mechanism and the position nested structure. The spatial memory encoding module extracts the spatial correlation characteristics of wind disturbances based on the position mapping and the disturbance distribution weighting mechanism.

[0066] The temporal trend features and the spatial correlation features are subjected to feature fusion and residual connection operations in the fusion control unit of the transformer prediction model to form a unified temporal-spatial coupling feature tensor; it should be noted that the fusion control unit of the transformer prediction model performs dimension alignment and attention weighted fusion operations between the temporal trend features and the spatial correlation features, calculates the interactive response relationship between the two types of features under the same position encoding, and forms a fused joint feature representation, and then introduces a residual connection mechanism to weightedly superimpose the fusion results and the original features, so as to maintain the original feature structure while enhancing the joint representation capability, and finally output a unified temporal-spatial coupling feature tensor;

[0067] The time-space coupling feature tensor is input into the decoding module of the converter prediction model. According to the preset target time domain sliding window, the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend in the future target time period are predicted, and the wind disturbance change trend prediction result is output;

[0068] Based on the wind disturbance change trend prediction results and the historical disturbance trajectory distribution in the spatial disturbance distribution map, the offset angle, offset direction and offset amplitude of the spray trajectory in the target time domain are calculated, and the spray trajectory offset parameters are output.

[0069] The decoding module of the transformer prediction model includes a position encoding unit, a multi-head attention mechanism layer, a feedforward neural network layer and an output mapping layer. The decoding module receives the time-space coupling feature tensor output by the fusion control unit, and combines the time index position of the current observed time step. The position encoding unit encodes the position index of each predicted time step in the target time period, generates the corresponding time position code, and embeds the generated time position code into the time-space coupling feature tensor to maintain the continuity of the time structure and the consistent transmission of the position dependency during the prediction process;

[0070] The multi-head attention mechanism layer in the decoding module is used to construct an attention mapping relationship between historical features and predicted positions at the target time position, extract contextual feature representations between historical time steps corresponding to each predicted time step with weight values ​​within a preset upper threshold range, and use the feedforward neural network layer to perform nonlinear transformations on the contextual feature representations to output prediction intermediate results with time-recursive characteristics.

[0071] The output mapping layer of the decoding module projects the intermediate prediction results into the prediction component spaces of wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate, forming the wind disturbance change trend prediction results;

[0072] Among them, the preset target time domain sliding window slides backward from the current moment by a preset time length, and the preset time length corresponds to the number of future time steps set in the converter prediction model; the decoding module performs attention weight allocation operations on the fused time-space coupling feature tensor in sequence at each prediction time step within the target time domain sliding window, and inputs the attention weighted results into the feedforward neural network layer for feature mapping processing to generate the disturbance prediction output corresponding to the time step, and gradually generates a sequence of predicted values ​​of each disturbance variable within the target time period, thereby obtaining the output of the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend; in addition, the preset target time domain sliding window refers to a fixed-length time interval extending backward from the current observation time point, which is used to determine the time range for the decoding module to perform continuous predictions in future time steps.

[0073] High-pressure atomizing nozzles in a local area where the wind speed fluctuation rate exceeds a preset wind speed fluctuation threshold, or high-pressure atomizing nozzles in a local area where the wind speed derivative change rate exceeds a preset wind speed derivative change threshold are defined as partial high-pressure atomizing nozzles;

[0074] During the non-dust reduction operation phase, based on the preset disturbance intensity sorting rules, a small amount of droplets for testing is released. The spraying of the small amount of droplets adopts a short-term intermittent spraying method with a duration of less than 1 second and a spray pressure lower than 50% of the lower limit of the conventional operating pressure, so as to ensure that the droplet distribution is controllable during the test and does not interfere with the normal dust reduction operation; the disturbance intensity sorting rules are based on the standard values ​​of the preset wind speed fluctuation rate and the wind speed derivative change rate, and define the linear combination formula of the two: disturbance intensity value = α × wind speed fluctuation rate + β × wind speed derivative change rate, where α and β are fixed weight coefficients, and the calculated disturbance intensity values ​​are sorted from large to small to determine the sprinkler response priority; the duration of less than 1 second and the spray pressure lower than 50% of the lower limit of the conventional operating pressure are used to avoid the formation of large-scale fog cluster diffusion, so that the micro-droplet spraying is only used for disturbance path testing without interfering with the normal spray dust reduction operation;

[0075] The image or sound wave return device includes an image recognition device or a sound wave positioning device, which is installed on an automatic lifting rod to collect the movement process of the trace droplets and obtain the drift path of the trace droplets based on continuous frame matching and position inversion methods; the drift path is subjected to path overlap analysis with the spray trajectory offset parameter output by the prediction module, and an offset error between the drift path and the predicted trajectory is constructed to characterize the relative error between the predicted result and the actual disturbance response;

[0076] Based on the offset error, error correction is performed on the offset angle, offset direction and offset amplitude in the spray trajectory offset parameters in turn. The error correction includes spatially decomposing the offset error between the drift path and the predicted trajectory into corresponding offset direction angle and offset amplitude components, and adjusting the corresponding parameters respectively; at the same time, based on the time difference between the response moment of the trace droplet drift path and the timestamp corresponding to the predicted trajectory, the time offset is calculated, and the time offset is used as a reference for evaluating the prediction hysteresis; based on the offset direction angle and offset amplitude components, the corrected offset angle, offset direction and offset amplitude results are fed back to the compensation module for real-time updating of the subsequent spray compensation control strategy, thereby constructing a dynamic correction mechanism that guides the adaptive adjustment of control parameters based on prediction deviation.

[0077] It should be generally explained that the formation process of this scheme is based on the problem background of drastic changes in the wind disturbance environment during the mobile operation of the spraying vehicle. By building a complete set of intelligent control spray dust reduction system, the interference problem of wind disturbance environment on spray trajectory control and dust reduction accuracy is systematically solved; its formation path starts from the real-time collection of wind field parameters at different heights and directions by the wind speed and direction sensors on the spraying platform. These sensor nodes are installed on the automatic lifting rod. As the first layer of information source for wind field perception, a wind speed gradient data set containing spatial structure is constructed; then the feature extraction module standardizes and features the data set Extraction operation, in which normalization processing unifies the scale of wind speed and direction data through mean shift and range scaling to ensure the comparability of data from different measurement points and time periods. Disturbance characteristics are then extracted by calculating wind speed vector differences and wind direction rotation offsets. A wind disturbance feature vector group is constructed through statistics of wind speed fluctuation rate, wind direction rotation frequency, and wind speed derivative change rate. A wind disturbance time series is then constructed through time sequence combination. A wind disturbance intensity gradient field and rotation direction distribution map are generated through spatial interpolation. Finally, a comprehensive spatial disturbance distribution map is formed, realizing a structured conversion from wind speed gradient data to disturbance spatiotemporal expression.

[0078] In the modeling and prediction stage, the scheme introduces a time-space dual memory nested transformer prediction model. The transformer prediction model takes the wind disturbance time series and the spatial disturbance distribution map as joint input, constructs the time input tensor and the space input tensor respectively, and inputs them into the time memory encoding module and the space memory encoding module of the model. The core features between the disturbance trend and the spatial structure are extracted through the attention mechanism, and then the attention weighted fusion and residual connection operations are performed by the fusion control unit to form a unified time-space coupled feature tensor, ensuring that the time evolution pattern and the spatial disturbance feature jointly drive the prediction logic; the decoding module uses position encoding and feature attention mechanism to extract context dependencies for each future time step in the target time domain sliding window and output the trend prediction results of wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate at continuous time points, and then combines the historical disturbance trajectory to complete the calculation of the spray trajectory offset parameters, providing a basis for subsequent calibration and compensation control;

[0079] After the prediction is complete, the calibration module releases a small amount of test droplets, tracks their drift path through image recognition or sonar positioning, and overlaps and analyzes the actual drift path with the predicted spray trajectory offset parameters to form a spatial offset error. The offset error is further decomposed into an offset direction angle and an offset amplitude, which are used to correct the original offset parameters. At the same time, the system also records the time difference between the actual occurrence of the drift response and the predicted time point to calculate the time offset, which serves as a reference indicator for measuring the prediction lag. The corrected offset angle, offset direction, and offset amplitude are finally fed back to the compensation module, driving it to generate multi-dimensional control instructions in real time, including attitude adjustment, spray angle correction, and spray pressure control, thereby completing the adaptive adjustment of the spray control parameters. The entire feedback chain forms a closed-loop control mechanism of prediction-verification-compensation-feedback. The closed-loop control module ensures that the system re-collects new wind field data and re-enters the complete process after each round of compensation control, thereby achieving continuous adaptation and self-optimization of the system in a dynamic wind disturbance environment.

[0080] The design adopts a modular structure consisting of wind field acquisition, feature modeling, spatiotemporal prediction, real-time verification and closed-loop control in order to ensure that the system has anti-disturbance and execution capabilities; the introduction of three types of disturbance characteristics, namely wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate, covers the amplitude change, directional instability and mutation trend of wind field disturbances, and improves the accuracy of wind disturbance modeling; and the time-space dual memory structure enables the model to predict mutation points and trend turning points by collaboratively modeling the time recursion and spatial propagation characteristics of disturbances; in the verification stage, a short-term, low-pressure intermittent spray strategy is adopted, and trace droplet tests are triggered only in the areas at the top of the disturbance intensity ranking to ensure that the test process is accurate and controllable and does not affect dust reduction operations; finally, through high-frequency feedback correction and iterative optimization of compensation instructions, the system achieves stable control of the spray trajectory under complex wind disturbance conditions, improves dust reduction efficiency and operation accuracy, and has engineering application prospects.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A mobile intelligent control spray dust suppression system, comprising an intelligent control system, an environmental treatment system, and a spray vehicle. The intelligent control system and the environmental treatment system are installed on the spray vehicle. The environmental treatment system includes an atomizing nozzle, a water storage tank, a power supply system, and a lifting rod. The system is characterized by: The intelligent control system includes a wind farm construction module, a feature extraction module, a prediction module, a calibration module, and a compensation module; The wind field construction module is used to collect local wind field parameters at different heights and directions using multiple wind speed and direction sensor nodes installed on the lifting mast while the spray dust suppression sprayer is moving, and to construct a wind speed gradient dataset based on the local wind field parameters; The feature extraction module is used to perform normalization and disturbance feature extraction processing on the wind speed gradient data set to generate a wind disturbance time series and a spatial disturbance distribution map; The prediction module is used to input the wind disturbance time series and spatial disturbance distribution map into the preset time-space dual memory nested transformer prediction model, and output the wind disturbance change trend prediction results and spray trajectory offset parameters in the target time domain; The calibration module is used to control some atomizing nozzles to release droplets for testing wind disturbance. It also monitors the drift path of the droplets based on images or acoustic wave feedback devices, compares the drift path with the spray trajectory offset parameters, and updates the spray trajectory offset parameters. The compensation module is used to generate control instructions for the atomizing nozzle's posture adjustment, spray angle correction, and spray pressure change based on the corrected spray trajectory offset parameters, and drives the environmental management system through the intelligent control system to complete the spray compensation control; In the prediction module, based on the wind disturbance time series and spatial disturbance distribution map generated by the feature extraction module, the wind speed fluctuation rate, wind direction rotation frequency, and wind speed derivative change rate are combined into a time input tensor at continuous time steps. At the same time, the wind disturbance intensity gradient field and rotation direction distribution map are combined into a spatial input tensor to construct a joint input structure for the converter prediction model. The time input tensor and the space input tensor are respectively input into the time memory encoding module and the space memory encoding module of the converter prediction model. The time memory encoding module extracts the temporal trend characteristics of wind disturbance changes based on the multi-head attention mechanism and the position nested structure. The spatial memory encoding module extracts the spatial correlation characteristics of wind disturbances based on the position mapping and the disturbance distribution weighting mechanism. Performing feature fusion and residual connection operations on the temporal trend features and the spatial correlation features in a fusion control unit of a transformer prediction model to form a unified temporal-spatial coupling feature tensor; The time-space coupling feature tensor is input into the decoding module of the converter prediction model. According to the preset target time domain sliding window, the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend in the future target time period are predicted, and the wind disturbance change trend prediction result is output; Based on the wind disturbance change trend prediction results and the historical disturbance trajectory distribution in the spatial disturbance distribution map, the offset angle, offset direction and offset amplitude of the spray trajectory in the target time domain are calculated, and the spray trajectory offset parameters are output; The decoding module of the transformer prediction model includes a position encoding unit, a multi-head attention mechanism layer, a feedforward neural network layer and an output mapping layer. The decoding module receives the time-space coupling feature tensor output by the fusion control unit, and combines the time index position of the current observed time step, encodes the position index of each predicted time step in the target time period through the position encoding unit, generates the corresponding time position code, and embeds the generated time position code into the time-space coupling feature tensor; The multi-head attention mechanism layer in the decoding module is used to construct an attention mapping relationship between historical features and predicted positions at the target time position, extract contextual feature representations between historical time steps corresponding to each predicted time step with weight values ​​within a preset upper threshold range, and use the feedforward neural network layer to perform nonlinear transformations on the contextual feature representations to output prediction intermediate results with time-recursive characteristics. The output mapping layer of the decoding module projects the intermediate prediction results into the prediction component spaces of wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate, forming the wind disturbance change trend prediction results; Among them, the preset target time domain sliding window slides backward from the current moment by a preset time length, and the preset time length corresponds to the number of future time steps set in the converter prediction model; the decoding module sequentially performs an attention weight allocation operation on the fused time-space coupling feature tensor at each prediction time step within the target time domain sliding window, and inputs the attention weighted result into the feedforward neural network layer for feature mapping processing to generate a disturbance prediction output corresponding to the time step, and gradually generates a sequence of predicted values ​​of each disturbance variable within the target time period, thereby obtaining the output of the wind speed fluctuation rate change trend, the wind direction rotation frequency change trend and the wind speed derivative change rate change trend; Atomizing nozzles in a local area where the wind speed fluctuation rate exceeds a preset wind speed fluctuation threshold, or atomizing nozzles in a local area where the wind speed derivative change rate exceeds a preset wind speed derivative change threshold are defined as partial atomizing nozzles; During the non-dust suppression operation phase, test droplets are released based on the preset disturbance intensity sorting rules. The spray of droplets adopts an intermittent spraying method with a duration of less than 1 second and a spray pressure lower than 50% of the lower limit of the normal operating pressure; The image or sound wave return device includes an image recognition device or a sound wave positioning device, which is installed on a lifting rod to collect the movement process of the droplets and obtain the drift path of the droplets based on continuous frame matching and position inversion methods; the drift path is subjected to path overlap analysis with the spray trajectory offset parameter output by the prediction module, and the offset error between the drift path and the predicted trajectory is constructed to characterize the relative error between the predicted result and the actual disturbance response; Based on the offset error, error correction is performed on the offset angle, offset direction and offset amplitude in the spray trajectory offset parameters in turn. The error correction includes spatially decomposing the offset error between the drift path and the predicted trajectory into corresponding offset direction angle and offset amplitude components, and adjusting the corresponding parameters respectively; at the same time, based on the time difference between the response moment of the droplet drift path and the timestamp corresponding to the predicted trajectory, the time offset is calculated, and the time offset is used as a reference for evaluating the prediction hysteresis.

2. The mobile intelligent spray dust suppression system according to claim 1, characterized in that: The intelligent control system also includes a closed-loop control module, which is used to re-collect the local wind field parameters of each wind speed and direction sensor node on the lifting rod after the compensation module completes the spray compensation control, and construct a new wind speed gradient data set; input the new wind speed gradient data set into the feature extraction module, and pass it through the feature extraction module, prediction module, calibration module and compensation module in sequence; through the periodic linkage call of the intelligent control system, a closed-loop control mechanism of wind disturbance prediction and spray control is formed with changes in local wind field parameters as input and control instruction updates as output.

3. The mobile intelligent spray dust suppression system according to claim 2, characterized in that: The atomizing nozzle is installed on the atomizing bracket at the rear of the spraying vehicle. The atomizing nozzle is connected to the water storage tank through a liquid pipe. The power supply system is set at the chassis of the spraying vehicle and provides power output to the atomizing nozzle, lifting rod and related control circuits through a line distribution interface. The lifting rod is fixed to the bracket structure on the top of the spraying vehicle. A wind speed and direction sensor node is integrated in the lifting rod. At the same time, the wind speed and direction sensor node is used as a front-end acquisition device for local wind field parameters. The output data of the wind speed and direction sensor node is transmitted to the intelligent control system. The intelligent control system and the environmental management system establish a two-way connection through the control harness and the data communication module. The intelligent control system is used to receive the local wind field parameters of the lifting rod and send control instructions for spraying to the atomizing nozzle according to the prediction module, calibration module and compensation module.

4. The mobile intelligent spray dust suppression system according to claim 3, characterized in that: When the feature extraction module performs normalization and disturbance feature extraction processing on the wind speed gradient data set, the normalization processing includes performing mean shift and range scaling processing on the collected wind speed and wind direction parameters at each altitude and azimuth according to a preset time window, that is, forming a wind speed gradient matrix under a unified scale; After completing the normalization process, the feature extraction module calculates the wind speed vector difference and wind direction rotation offset between adjacent measuring points at the same time section based on the wind speed gradient matrix, and counts the wind speed fluctuation rate, wind direction rotation frequency and wind speed derivative change rate within the specified time step as wind disturbance features to form a wind disturbance feature vector group; The wind disturbance feature vector groups are arranged in chronological order to construct a wind disturbance time series. At the same time, the feature extraction module, based on the spatial position of each measuring point of the wind speed and direction sensor nodes, performs spatial interpolation on the wind disturbance features within the corresponding time window and maps them to the wind field area distribution map, thereby solving the local wind disturbance intensity gradient field and rotation direction distribution map to form a spatial disturbance distribution map.

5. The mobile intelligent spray dust suppression system according to claim 4, characterized in that: The wind speed fluctuation rate of the wind disturbance characteristic is calculated by calculating the ratio of the standard deviation to the mean of the wind speed values ​​at the same measuring point within a specified time window to reflect the relative change amplitude of the wind speed within the time window; The wind direction rotation frequency of the wind disturbance feature is measured by counting the number of times the wind direction changes exceed a preset angle threshold within a specified time window and normalizing it to the change frequency per unit time to measure the wind direction instability; The wind speed derivative change rate of the wind disturbance feature is calculated by calculating the difference value of the wind speed derivative in adjacent time steps and performing statistics on the derivative change in unit time to represent the sudden change intensity of the wind speed change trend.

Images