Particle concentration-based air supply air speed adaptive regulation method and system

Abstract

The application relates to the technical field of intelligent environment control, and discloses a particle concentration-based air supply wind speed self-adaptive regulation and control method and system, which comprises the following steps: obtaining a concentration time sequence, particle position coordinates, environmental interference factors, real-time wind speed, airflow flow field characteristics and external environmental load, performing heterogeneous characteristic analysis to obtain heterogeneous distribution characteristics; performing concentration trend curve fitting according to the heterogeneous distribution characteristics to obtain a concentration trend curve; performing difference region division according to the concentration trend curve to obtain a regional difference distribution; performing regional weight calculation and weight sorting according to the regional difference distribution to obtain regional priority sorting; performing configuration association matching according to the regional priority sorting to obtain ventilation configuration; performing efficiency verification according to the ventilation configuration to obtain an efficiency improvement degree; and performing efficiency verification solidification according to the efficiency improvement degree to obtain a continuous improvement path. The method can realize efficient management of industrial environment air quality.

Description

Technical Field

[0001] This invention relates to the field of intelligent environmental control technology, and in particular to a method and system for adaptive control of air supply speed based on particle concentration. Background Technology

[0002] Currently, air quality control in industrial and built environments is crucial to human health and energy utilization, necessitating efficient ventilation to address dynamic changes. The uneven spatial and temporal distribution of particle concentration, with concentrations significantly higher near pollution sources than in other areas, makes it difficult for industrial control systems to identify localized differences. Without dynamic adjustments based on these regional concentration variations, uneven ventilation and energy waste are likely to occur. Therefore, technologies capable of real-time sensing of the spatial distribution of particle concentration and dynamically adjusting airflow speeds in different zones are needed.

[0003] In one existing technology, discrete sensor nodes are deployed within the working area to collect particle concentration data. The real-time concentration value of each node is compared one by one with a preset global concentration threshold. When the concentration value of any node exceeds the threshold, the ventilation system is triggered to operate at a constant airflow. The system periodically checks the concentration value of each node at preset time intervals. If the detected concentration value remains above the threshold, the current airflow speed is maintained. If the concentration values ​​of all nodes fall below the threshold, the ventilation system is shut down or switched to a low-speed standby mode until the next trigger. This scheme adjusts the global airflow speed uniformly based on the triggering status of individual nodes. However, existing technologies that trigger unified airflow based on a global threshold can lead to localized concentration exceedances causing global over-ventilation, resulting in energy waste and uneven air quality.

[0004] Therefore, existing technologies cannot perform differentiated regulation based on the spatial distribution of concentration. Summary of the Invention

[0005] This invention provides a method and system for adaptive control of air supply speed based on particle concentration, so as to achieve efficient management of air quality in industrial environments.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for adaptive control of air supply velocity based on particle concentration, comprising:

[0007] Acquire concentration time series, particle position coordinates, environmental disturbance factors, real-time wind speed, airflow field characteristics, and external environmental load;

[0008] Based on the concentration time series and the particle position coordinates, heterogeneous feature analysis is performed by fitting node evolution trajectories to obtain heterogeneous distribution characteristics.

[0009] Based on the heterogeneous distribution characteristics and the concentration time series, the fluctuation amplitude is compared to obtain the peak period, and the concentration trend curve is fitted based on the peak period to obtain the concentration trend curve;

[0010] Based on the concentration trend curve and the environmental disturbance factor, the unit concentration is corrected to obtain the corrected grid unit. Based on the corrected grid unit, the difference region is divided according to the concentration difference comparison to obtain the regional difference distribution.

[0011] Based on the regional difference distribution, state classification is performed based on delay judgment to obtain a stable state, and regional weights are calculated and sorted according to the stable state to obtain a regional priority ranking.

[0012] Based on the regional priority ranking, the airflow field characteristics, and the real-time wind speed, a wind speed consistency check is performed to obtain an accuracy coefficient, and a configuration association matching is performed to obtain a ventilation configuration.

[0013] Based on the ventilation configuration and the external environmental load, hysteresis parameters are optimized to obtain an updated parameter set, and performance is verified to determine the degree of efficiency improvement.

[0014] Based on the degree of efficiency improvement and the concentration time series, deviation-driven parameters are adjusted to generate optimized control commands, and performance is verified and solidified to obtain a continuous improvement path.

[0015] Secondly, the present invention provides an adaptive control system for air supply velocity based on particle concentration, comprising:

[0016] The data acquisition module is used to acquire concentration time series, particle position coordinates, environmental interference factors, real-time wind speed, airflow field characteristics, and external environmental loads;

[0017] The feature analysis module is used to perform heterogeneous feature analysis by fitting node evolution trajectories based on the concentration time series and the particle position coordinates to obtain heterogeneous distribution features;

[0018] The curve fitting module is used to compare the fluctuation amplitude based on the heterogeneous distribution characteristics and the concentration time series to obtain the peak period, and to fit the concentration trend curve based on the peak period to obtain the concentration trend curve.

[0019] The region division module is used to correct the unit concentration based on the concentration trend curve and the environmental interference factor to obtain corrected grid units, and to divide the difference regions based on the concentration difference comparison of the corrected grid units to obtain the regional difference distribution.

[0020] The weight calculation module is used to classify states based on the regional difference distribution and delay judgment to obtain stable states, and to calculate and sort regional weights based on the stable states to obtain regional priority ranking.

[0021] The configuration matching module is used to perform wind speed consistency verification based on the area priority sorting, the airflow field characteristics and the real-time wind speed, obtain the accuracy coefficient, and perform configuration association matching to obtain the ventilation configuration;

[0022] The performance verification module is used to optimize the hysteresis parameters based on the ventilation configuration and the external environmental load, obtain an updated parameter set, and perform performance verification to obtain the degree of efficiency improvement.

[0023] The output module is used to adjust the deviation driving parameters according to the degree of efficiency improvement and the concentration time series, generate optimized control instructions, and perform performance verification and solidification to obtain a continuous improvement path.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] (1) This invention constructs the spatial heterogeneous distribution characteristics of particle concentration by extracting local extreme points in the diffusion gradient distribution and tracing their evolution trajectory. Since the local extreme points accurately locate the pollution source and the stagnation point where the airflow converges, the evolution trajectory reveals the migration path of the pollution cloud in three-dimensional space, enabling the spatial heterogeneous distribution characteristics to truly reflect the non-uniformity and dynamic change law of the concentration distribution. This solves the problem of ignoring local differences and global unified control in the existing technology, and provides a precise basis for zoned differentiated air supply.

[0026] (2) This invention constructs a stability evaluation model for the control system by integrating the variance analysis of the refresh frequency sequence and the median filtering of the feedback delay sequence, and performs linear weighted summation based on the regional difference distribution to achieve regional priority ranking. Since the refresh frequency variance quantifies the communication stability of the data link, and the feedback delay sequence reflects the true response capability of the actuator after median filtering, the integration of the two with the urgency of pollution can take into account both purification needs and control safety, and solves the defects of the existing technology that only triggers regulation based on the concentration threshold and ignores the stability of the control system itself.

[0027] (3) Based on the response delay sequence and external environmental load, the present invention constructs an adaptive curve, identifies the lag interval, dynamically increases the acceleration threshold of the servo motor, and verifies the optimization effect by reducing the convergence time ratio, thus forming a closed-loop performance verification mechanism. Since the adaptive curve reveals the correlation between response delay and external disturbance, the lag interval locates the working condition where the actuator response capability is insufficient, and the acceleration threshold increase enhances the instantaneous torque output capability, solving the problem of lack of parameter adaptive adjustment and effect verification in the prior art, and realizing the continuous improvement of the response efficiency of the ventilation system. Attached Figure Description

[0028] Figure 1This is a schematic diagram of the adaptive control method for air supply speed based on particle concentration provided in the first embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of the adaptive control system for air supply speed based on particle concentration provided in the second embodiment of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Reference Figure 1 The first embodiment of the present invention provides an adaptive control method for air supply velocity based on particle concentration, comprising the following steps:

[0032] S11, acquire concentration time series, particle position coordinates, environmental interference factors, real-time wind speed, airflow field characteristics and external environmental load;

[0033] S12, Based on the concentration time series and the particle position coordinates, heterogeneous feature analysis is performed by fitting the node evolution trajectory to obtain heterogeneous distribution characteristics;

[0034] S13. Based on the heterogeneous distribution characteristics and the concentration time series, the fluctuation amplitude is compared to obtain the peak period, and the concentration trend curve is fitted based on the peak period to obtain the concentration trend curve.

[0035] S14. Based on the concentration trend curve and the environmental disturbance factor, the unit concentration is corrected to obtain the corrected grid unit. Based on the corrected grid unit, the difference region is divided according to the concentration difference comparison to obtain the regional difference distribution.

[0036] S15, based on the regional difference distribution, state classification is performed based on delay judgment to obtain a stable state, and regional weight calculation and weight sorting are performed based on the stable state to obtain regional priority sorting.

[0037] S16. Based on the regional priority sorting, the airflow field characteristics, and the real-time wind speed, perform wind speed consistency verification to obtain an accuracy coefficient, and perform configuration association matching to obtain a ventilation configuration.

[0038] S17. Based on the ventilation configuration and the external environmental load, perform hysteresis parameter optimization to obtain an updated parameter set, and perform performance verification to obtain the degree of efficiency improvement.

[0039] S18, based on the degree of efficiency improvement and the concentration time series, the deviation driving parameters are adjusted, an optimized control command is generated, and the performance is verified and solidified to obtain a continuous improvement path.

[0040] In step S11, the concentration time series, particle position coordinates, environmental interference factors, real-time wind speed, airflow field characteristics, and external environmental load are obtained.

[0041] Specifically, the system acquires real-time concentration time series data from a high-density sensor network deployed in the industrial environment. This time series includes particle concentration values ​​collected by each sensor node at continuous sampling times, along with the corresponding sampling timestamps. Each sensor node has a unique identifier. Particle position coordinates are obtained from the configuration file during system initialization. These coordinates record the precise position of each sensor node in three-dimensional space, determining the spatial topology between nodes. Environmental interference factors are acquired in real-time from wind speed, humidity, and electrostatic field sensors deployed in the same industrial environment. These factors include the local wind speed, relative humidity, and electrostatic field strength at the location of each sensor node. Real-time wind speed is acquired from the ventilation system controller, representing the current airflow velocity value fed back by each air supply terminal device. Airflow velocity and direction data are collected in real-time from airflow sensors deployed at air inlets and outlets. Combined with a pre-set computational fluid dynamics model, airflow field characteristics are analyzed to describe the airflow pattern category in each area. External environmental loads, including outdoor wind pressure changes and air supply temperature differences, are acquired in real-time from wind pressure sensors deployed outside the industrial environment and temperature sensors inside the air supply ducts.

[0042] All acquired data are aligned using timestamps as indices to provide input for subsequent steps. Concentration time series, particle location coordinates, and environmental interference factors are used to construct a spatial distribution model of particle concentration. Real-time wind speed and airflow field characteristics are used to assess the current operating status of the ventilation system and airflow patterns. External environmental loads are used to analyze the impact of external disturbances on the ventilation system. The sampling frequency of the concentration time series is preset to a fixed value, set to once per second based on the sensor hardware's maximum sampling capability, to ensure the capture of rapid changes in particle concentration. Local wind speed data from the environmental interference factors are used in subsequent steps to calculate interference weights. A preset weight mapping table converts wind speed values ​​into correction coefficients. This weight mapping table, based on calibration data provided by the sensor manufacturer, divides wind speed into several intervals, each corresponding to a preset weight value. Airflow field characteristics are obtained by matching a computational fluid dynamics model with real-time airflow data. This model is pre-constructed based on the geometric structure of the industrial environment, the layout of air inlets and outlets, and compares the real-time collected airflow velocity vectors with the flow field patterns in the model to identify the airflow pattern category in the current area. The outdoor wind pressure change value in the external environmental load is collected in real time by the wind pressure sensor, and the supply air temperature difference value is calculated by the difference between the supply air temperature sensor and the indoor temperature sensor.

[0043] It is worth noting that the computational fluid dynamics model uses a three-dimensional geometric model of the industrial environment as the computational domain, employs unstructured mesh generation, sets the air inlet as the velocity inlet boundary and the exhaust outlet as the pressure outlet boundary, solves the Navier-Stokes equations and continuity equations for incompressible fluids, and adopts the standard model for turbulence. The reference airflow velocity vector field at each location is obtained through steady-state solution. During runtime, the real-time collected airflow velocity vector is matched with the reference field using cosine similarity. When the similarity exceeds 80%, it is determined to be the corresponding airflow pattern.

[0044] In step S12, based on the concentration time series and the particle position coordinates, heterogeneous feature analysis is performed through node evolution trajectory fitting to obtain heterogeneous distribution features, including:

[0045] Based on the concentration time series, the particle position coordinates, and the preset sampling frequency, the concentration change rate is obtained by calculating the particle concentration difference at adjacent sampling times for the same node.

[0046] Based on the concentration change rate, the diffusion gradient distribution is obtained by calculating the ratio of the rate difference between adjacent nodes to the spatial distance.

[0047] Calculate the difference between the diffusion gradient of each node and the diffusion gradient of its neighboring nodes in the diffusion gradient distribution to obtain the gradient difference, and filter out the nodes corresponding to the gradient difference exceeding a preset jump threshold to obtain local extreme points.

[0048] Based on the local extreme points, cross-time node association is performed using the K nearest neighbor algorithm to obtain a sequence of common-origin nodes. Based on the sequence of common-origin nodes, curve fitting is performed using cubic spline interpolation to obtain a continuous evolution trajectory.

[0049] The starting position, direction of movement, and spatial coverage of the continuous evolution trajectory are extracted to obtain the heterogeneous distribution characteristics.

[0050] Specifically, based on the concentration time series, particle position coordinates, and a preset sampling frequency, for each sensor node, the particle concentration values ​​at two adjacent sampling times are extracted. The concentration value at the later time is subtracted from the concentration value at the previous time to obtain the concentration change rate of that node within that time interval. The preset sampling frequency is set according to the maximum sampling capability of the sensor hardware, for example, once per second, to ensure that the time interval between adjacent sampling times is fixed at 1 second.

[0051] Based on the concentration change rate and particle position coordinates of each sensor node, a sensor node topology network is constructed, connecting each node to its spatially adjacent nodes. For each pair of adjacent nodes, the difference in concentration change rates between the two nodes is calculated, and the spatial distance between them is calculated based on their particle position coordinates. The difference is divided by the spatial distance to obtain the diffusion gradient value between the adjacent node pair. This process is repeated for all adjacent node pairs, defining the diffusion gradient value of each node as the average of its diffusion gradient values ​​with those of all its adjacent nodes, ultimately forming a diffusion gradient distribution covering the entire monitoring area.

[0052] Based on the diffusion gradient distribution, for each sensor node, the difference between its diffusion gradient value and the diffusion gradient values ​​of all its neighboring nodes is calculated. The maximum absolute value of all differences is taken as the gradient difference for that node. This gradient difference is compared to a preset jump threshold. If the gradient difference exceeds the jump threshold, the node is marked as a local extremum. The jump threshold is determined through historical data statistics. Gradient difference data from all nodes are collected during a preset period of continuous operation under normal system conditions. The percentiles of these data are calculated, and the 95th percentile is taken as the jump threshold, ensuring that a local extremum is only marked when the gradient change is extreme. Local extrema correspond to the location of pollution sources or points where airflow converges and stagnates.

[0053] Based on local extrema marked by multiple consecutive sampling times, the K-nearest neighbor algorithm is used to associate nodes across time points, with K set to 1, meaning only the spatially closest points are matched. For each local extrema at the current time, its spatial distance to all local extrema at the previous time is calculated. The extrema at the previous time with the smallest distance that is below a preset distance threshold is selected, and these two extrema are determined to be the locations of the same pollution source at different times, forming a sequence of homologous nodes. The preset distance threshold is determined based on the sensor node deployment spacing, taking 1.5 times the average node spacing to ensure that the displacement of the same pollution source at adjacent time points does not exceed this range. For successfully matched homologous node sequences, they are arranged in chronological order, and cubic spline interpolation is used for curve fitting to convert discrete node positions into continuous curves, obtaining a continuous evolution trajectory. The cubic spline interpolation constructs a cubic polynomial between each pair of adjacent nodes, ensuring the continuity of the first and second derivatives at the nodes, making the trajectory smooth and accurately passing through all node positions.

[0054] Based on the continuous evolution trajectory, the coordinates of the earliest node in the trajectory are extracted as the starting position. The overall direction vector of the trajectory is calculated as the direction of movement. The coordinate ranges of the trajectory point set in the horizontal, vertical, and longitudinal directions are calculated separately. These three ranges are multiplied together and then multiplied by the total length of the trajectory to obtain the spatial coverage area. The starting position corresponds to the approximate location of the pollution source, the direction of movement reflects the dominant direction of pollutant diffusion, and the spatial coverage area delineates the high-concentration influence zone. The starting position, direction of movement, and spatial coverage area are integrated into a heterogeneous distribution characteristic.

[0055] In step S13, based on the heterogeneous distribution characteristics and the concentration time series, the fluctuation amplitude is compared to obtain the peak period, and a concentration trend curve is fitted based on the peak period to obtain the concentration trend curve, including:

[0056] Based on the heterogeneous distribution characteristics, the concentration time series is filtered and denoised using a weighted moving average algorithm to obtain a smoothed series.

[0057] The difference between local maxima and local minima in the smoothed sequence is calculated to obtain the fluctuation amplitude;

[0058] The peak time period is obtained by filtering out the time segments in the smoothed sequence where the fluctuation amplitude exceeds a preset fluctuation amplitude threshold;

[0059] Polynomial fitting was performed on the concentration values ​​during the peak period to obtain the concentration trend curve.

[0060] Specifically, based on the high-concentration influence areas indicated by the starting position, direction of movement, and spatial coverage in the heterogeneous distribution characteristics, the data of the corresponding areas in the concentration time series are extracted as the sequence to be processed. A weighted moving average algorithm is used to filter and denoise this sequence. The sliding window length is set to a preset value, which is determined based on the sampling frequency. The number of sampling points within the time window corresponds to the sampling frequency; for example, if the sampling frequency is once per second and the window duration is 5 seconds, then the window length is 5. For each sampling point within the window, a weight is assigned based on the relative distance between the spatial coverage area in the heterogeneous distribution characteristics and the current point's location. Sampling points closer to the center of the high-concentration influence area are assigned higher weights, and those farther away are assigned lower weights. The sum of the weights is normalized to 1. The concentration values ​​of each sampling point within the window are multiplied by their corresponding weights and summed to obtain the smoothed value at the center of the window. The smoothed sequence is obtained by traversing the entire sequence through the sliding window.

[0061] It should be noted that the weight coefficients in the weighted moving average algorithm are dynamically set based on the heterogeneous distribution characteristics. These characteristics include the starting position, direction of movement, and spatial coverage, where the spatial coverage defines the high-concentration influence area. For each sampling point in the smoothing sequence, the spatial distance between the node and the starting position is calculated based on the particle position coordinates of the sensor node to which the sampling point belongs. This distance is then compared with the spatial coverage to determine whether the node is within or outside the high-concentration influence area. If the node is within the high-concentration influence area, the concentration value at the current sampling time is assigned a higher weight, while the concentration values ​​of historical sampling points are assigned lower weights. Specifically, the weight at the current time is set to a first preset value, and the weights of historical sampling points are set to a second preset value, where the first preset value is greater than the second preset value, and the sum of all weights is normalized to 1. If the node is outside the high-concentration influence area, the concentration value at the current sampling time is assigned a lower weight, while the concentration values ​​of historical sampling points are assigned a higher weight to enhance the smoothing effect. Specifically, the weight at the current time is set to the second preset value, and the weights of historical sampling points are set to the first preset value, where the first preset value is greater than the second preset value. The ratio of the first preset value to the second preset value is determined based on the ratio of the spatial coverage area to the spacing between sensor node deployments. The larger this ratio, the greater the difference between the first and second preset values, thus enhancing the differentiation in weight allocation between areas within and outside the region. Through this method, the weighted moving average algorithm can dynamically adjust the filtering intensity based on the spatial heterogeneity of particle concentration distribution, retaining more real-time information in areas with drastic high concentration changes and enhancing noise suppression capabilities in areas with stable concentrations.

[0062] Based on the smoothed sequence, local maxima and local minima are identified using the sliding window method. During the identification process, for each sampling point, its concentration value needs to be compared with the concentration values ​​of the preset number of sampling points before and after it. This preset number is determined through autocorrelation analysis. Specifically, during the system initialization phase, historical concentration time series are collected within a preset duration of continuous operation under stable conditions. For each offset, the original sequence and the offset sequence are multiplied point by point, summed, and then divided by the sequence variance to obtain a curve showing the correlation coefficient changing with the offset. The offset corresponding to the first time the correlation coefficient falls below a preset correlation threshold is taken as the window length. The preset correlation threshold is 0.5, indicating that when the correlation between the offset sequence and the original sequence is lower than a moderate correlation, this time interval is considered sufficient to distinguish adjacent fluctuation characteristics. The window length is divided by the sampling interval and rounded up to obtain the preset number. If the concentration value of the current point is greater than the concentration values ​​of the preset number of sampling points before and after it, it is marked as a local maximum; if the concentration value of the current point is less than the concentration values ​​of the preset number of sampling points before and after it, it is marked as a local minimum. For each pair of adjacent local maxima and local minima, calculate the absolute value of the difference between their concentration values ​​to obtain the fluctuation range within that interval. After traversing all pairs of extreme points, a fluctuation range sequence is obtained.

[0063] Each fluctuation amplitude in the fluctuation amplitude sequence is compared with a preset fluctuation amplitude threshold, which is determined through historical data statistics. All fluctuation amplitude data are collected continuously for a preset duration under normal operating conditions. The percentiles of these data are calculated, and the 80th percentile is used as the fluctuation amplitude threshold to ensure that only fluctuations at a high level are considered valid fluctuations. Extreme value pairs where the fluctuation amplitude exceeds the fluctuation amplitude threshold are selected, and the corresponding time intervals are extracted to obtain the peak period.

[0064] Based on the peak period, smoothed sequence concentration values ​​for all sampling times within that period are extracted. These concentration values ​​are paired with their corresponding sampling times to form data point pairs. A least squares method is used for polynomial fitting, with the sampling time as the independent variable and the concentration value as the dependent variable. A polynomial function is sought that minimizes the sum of squares of the deviations between the function values ​​and the actual concentration values ​​at all sampling points. The polynomial order can be preset to 2nd or 3rd order depending on the complexity of the concentration fluctuations within the peak period to balance fitting accuracy and overfitting risk. The resulting continuous polynomial function plotted on the time axis is the concentration trend curve. This curve reflects the continuous trend of particle concentration changes over time within the peak period and can be used for concentration prediction and correction of subsequent grid cells.

[0065] In step S14, based on the concentration trend curve and the environmental disturbance factor, the unit concentration is corrected to obtain corrected grid cells. Then, based on the corrected grid cells and concentration difference comparisons, difference regions are divided to obtain the regional difference distribution, including:

[0066] Based on the concentration time series and the preset three-dimensional spatial coordinates, the concentration grid cells are obtained by dividing the grid using a uniform spatial grid division method.

[0067] When the slope of the tangent of the concentration trend curve exceeds the preset concentration change rate threshold, the interference weight is obtained by performing a weight lookup through a preset weight mapping table according to the environmental interference factor, and the concentration in the concentration grid cell is weighted and corrected according to the interference weight to obtain the corrected grid cell.

[0068] When the slope of the tangent line of the concentration trend curve does not exceed the concentration change rate threshold, the concentration grid cell is directly used as the correction grid cell;

[0069] Calculate the concentration difference between adjacent modified grid cells to obtain the concentration difference;

[0070] When the concentration difference is greater than the preset tolerance threshold, the current corrected grid cell is directly marked as an independent region to obtain the segmented region;

[0071] When the concentration difference is not greater than the tolerance threshold, the adjacent corrected grid cells are merged into the region where the current corrected grid cell is located, thus obtaining the segmented region;

[0072] Calculate the variance of the concentration gradient between each segmented region to obtain the regional difference distribution.

[0073] Specifically, a uniform spatial grid is created based on the concentration value at the current moment in the concentration time series and preset three-dimensional spatial coordinates. These preset three-dimensional spatial coordinates are determined through measurement and configuration during system initialization. During system deployment, a laser rangefinder is used to conduct on-site measurements of the monitoring environment, obtaining the minimum and maximum coordinate values ​​in the horizontal, vertical, and triangular directions, forming a three-dimensional spatial coordinate system containing six boundary values. Based on the preset three-dimensional spatial coordinates, the three-dimensional spatial region is divided equally in the horizontal, vertical, and triangular directions with a preset grid step size. The number of divisions in each direction is obtained by dividing the coordinate range by the grid step size and rounding up. Each grid cell is assigned a unique row, column, and layer index. For each sensor node, its corresponding grid cell is determined based on its particle position coordinates. The current concentration value of that node is assigned to that grid cell. If multiple sensor nodes are contained within the same grid cell, the arithmetic mean of their concentration values ​​is taken as the concentration value of that grid cell. Finally, a concentration grid cell containing both the concentration value and the spatial index is obtained.

[0074] The preset grid step size is determined based on the average spacing of the sensor nodes. It is half of the minimum horizontal spacing of all adjacent sensor nodes and half of the minimum height difference of each layer of sensor nodes in the vertical direction, ensuring that each grid cell contains at least one sensor node and that the total number of grid cells does not exceed twice the number of sensor nodes.

[0075] Based on the concentration trend curve, calculate the tangent slope at the current moment. The concentration trend curve is a continuous function with time as the independent variable and concentration as the dependent variable, and the tangent slope is the first derivative of this function on the time axis at the current moment. After obtaining the tangent slope, compare it with a preset concentration change rate threshold. The concentration change rate threshold is determined through historical data statistics. Collect concentration change rate data for all moments within a preset time period under normal operating conditions, calculate the percentiles of these data, and take the 90th percentile value as the concentration change rate threshold to ensure that the concentration correction operation is triggered only when the concentration change rate is at a high level.

[0076] When the slope of the tangent line of the concentration trend curve exceeds a preset concentration change rate threshold, a concentration correction operation is performed. Environmental interference factor data affecting each grid cell is acquired. These factors include local wind speed, relative humidity, and electrostatic field strength, and each factor corresponds to a preset weight mapping table. The weight mapping table is constructed during system initialization, dividing the measurement range of the environmental interference factors into several continuous intervals. Each interval corresponds to a weight value, determined based on the degree of influence of the interference factor on the concentration measurement deviation in the calibration data provided by the sensor manufacturer. The greater the deviation, the smaller the weight value, thus weakening the contribution of the interference factor to the concentration value. For each grid cell, based on the real-time collected environmental interference factor values, the weight value corresponding to each interference factor is queried, and the weight values ​​are multiplied to obtain the interference weight of the grid cell. The concentration value within the concentration grid cell is multiplied by the interference weight to obtain the corrected concentration value, forming the corrected grid cell. When the slope of the tangent line of the concentration trend curve does not exceed the preset concentration change rate threshold, the concentration grid cell is directly used as the corrected grid cell without weighted correction.

[0077] In one implementation, the weighted mapping table for relative humidity and electrostatic field strength is constructed in the same way as the weighted mapping table for local wind speed. Both are based on the degree of influence of the interference factor on the particle concentration measurement deviation in the calibration data provided by the sensor manufacturer. The measurement range is divided into three continuous intervals: low, medium, and high. Each interval corresponds to a weight value. The weight value for the interval with a deviation influence of less than 10% is 1.0, the weight value for the interval with a deviation influence of 10% to 30% is 0.8, and the weight value for the interval with a deviation influence of more than 30% is 0.6.

[0078] Based on the corrected grid cells, the concentration difference between adjacent grid cells is calculated. For each grid cell, its six adjacent grid cells are traversed, and the corrected concentration value of the current grid cell is subtracted from the corrected concentration value of the adjacent grid cells. The absolute value is then used to obtain the concentration difference in that direction. The concentration difference is compared with a preset tolerance threshold, which is determined through historical data statistics. Concentration difference data of all adjacent grid cells are collected during a preset period of continuous operation under normal system conditions. The percentiles of these data are calculated, and the 50th percentile value is taken as the tolerance threshold, serving as the dividing point for distinguishing between similar and dissimilar concentrations.

[0079] When the concentration difference exceeds a preset tolerance threshold, the current grid cell is marked as an independent region. When the concentration difference is not greater than the preset tolerance threshold, adjacent grid cells are merged into the region where the current grid cell is located; that is, two grid cells with similar concentrations are considered to be in the same region. This comparison and merging operation is repeated across all grid cells to obtain a segmented region composed of several connected regions.

[0080] Based on the segmented regions, calculate the concentration gradient difference between each region and its neighboring regions; for each segmented region, obtain the concentration gradient values ​​at all boundaries of the region, and calculate the gradient value of each boundary as described above. Calculate the variance of these boundary gradient values ​​as the regional difference coefficient of the region; associate the regional difference coefficient of each segmented region with its region identifier to form a regional difference distribution; this distribution reflects the degree of concentration change at the boundaries of different regions and is used for subsequent region priority ranking.

[0081] In step S15, based on the regional difference distribution, state classification is performed based on delay judgment to obtain a stable state, and regional weights are calculated and sorted according to the stable state to obtain a regional priority ranking, including:

[0082] Based on the regional difference distribution, the real-time refresh frequency and feedback delay of each segmented region are extracted to obtain a refresh frequency sequence and a feedback delay sequence.

[0083] When the variance of the refresh frequency sequence exceeds a preset frequency variance threshold, the feedback delay sequence is smoothed by a median filtering algorithm to obtain a standard delay sequence.

[0084] When the variance of the refresh frequency sequence does not exceed the frequency variance threshold, the feedback delay sequence is directly used as the standard delay sequence.

[0085] Based on the standard delay sequence, a stable state is obtained by classifying and identifying it using a pre-built state classification model.

[0086] Based on the stable state and the regional difference distribution, a linear weighted sum is performed using preset weight coefficients to obtain the regional ranking weights. The regional ranking weights are then sorted in descending order to obtain the regional priority ranking.

[0087] Specifically, based on the regional differences, for each segmented region, the refresh frequency and feedback delay of the corresponding air supply terminal device in that region are extracted in real time from the ventilation system controller to obtain a refresh frequency sequence and a feedback delay sequence. The refresh frequency refers to the frequency at which the air supply terminal device reports data to the controller, and the feedback delay refers to the time difference, measured in milliseconds, between the controller issuing a control command and the actuator completing the action and providing feedback. Each segmented region corresponds to one refresh frequency sequence and one feedback delay sequence. The sequence length is determined based on a preset monitoring duration, taking continuous data points from the most recent period. The preset monitoring duration is determined based on the response time of the controlled variable. During the system initialization phase, a step command is issued to the actuator, and the time required for the controlled variable to first reach and stabilize within the target range from its initial value is recorded. Twice this time is taken as the monitoring duration; if this time is less than ten seconds, the monitoring duration is twenty seconds.

[0088] The variance of the refresh frequency sequence is calculated based on the refresh frequency sequence. The calculation process involves calculating the arithmetic mean of all data in the refresh frequency sequence, then squared the difference between each data point and the mean, and finally calculating the arithmetic mean of these squared values ​​to obtain the refresh frequency variance. This refresh frequency variance is then compared to a preset frequency variance threshold. The frequency variance threshold is determined through historical data statistics. Refresh frequency variance data are collected for all moments within a preset duration of continuous operation under normal system conditions. The percentiles of these data points are calculated, and the 90th percentile value is used as the frequency variance threshold.

[0089] When the variance of the refresh rate sequence exceeds a preset frequency variance threshold, the feedback delay sequence is smoothed using a median filtering algorithm. The median filtering algorithm sets a sliding window length, determined by the sampling frequency, and selects the number of data points within the corresponding time window. The window length is preset to be an odd number. For each data point in the feedback delay sequence, a window is formed by that point and a preset number of data points before and after it. All data within the window are sorted by value, and the data in the middle position after sorting is taken as the output value of that point. The standard delay sequence is obtained after the sliding window traverses the entire sequence.

[0090] When the variance of the refresh frequency sequence does not exceed the preset frequency variance threshold, the feedback delay sequence is directly used as the standard delay sequence without filtering.

[0091] Based on the standard delay sequences, a pre-built state classification model is used to classify and identify stable states. Historical standard delay sequences and their corresponding historical stable states are obtained as training datasets. The historical standard delay sequences are feedback delay data recorded by the system during its historical operation, which are formed after median filtering. Each sequence corresponds to the delay value change trajectory of a segmented region within a continuous time window. The historical stable states are the system operating state categories at the corresponding moments, which are manually labeled or automatically determined by the system, including but not limited to steady-state response (delay value is stable within a low fluctuation range), damped oscillation (delay value exhibits gradually decaying periodic fluctuations), and divergent instability (delay value continuously increases or fluctuates violently). Each historical standard delay sequence is used as an input sample, and its corresponding stable state is used as the output label.

[0092] For each historical standard delayed sequence, the sequence mean, variance, maximum value, minimum value, mean of the absolute value of the first difference of the sequence, the autocorrelation coefficient of the sequence, such as lag 1 and lag 2, and the peak factor of the sequence are extracted. The above features constitute a multidimensional feature space for subsequent classification.

[0093] The state classification model is constructed using the Support Vector Machine (SVM) algorithm. A Radial Basis Function (RBF) kernel is used to map feature vectors to a high-dimensional space to construct the classification hyperplane. The model includes a penalty coefficient C and RBF kernel parameters γ. During model training, the training dataset is randomly divided into a 70% training set and a 30% test set, using stratified sampling to maintain a consistent proportion of each class. The feature vectors are standardized to ensure zero mean and unit variance for each feature dimension. A grid search method is used to search for the optimal combination of C and γ within a preset range. The search range for C is [0.1, 1, 10, 100, 1000], and the search range for γ is [0.001, 0.01, 0.1, 1, 10]. For each pair (C, γ), five-fold cross-validation is used to calculate the average classification accuracy. The pair (C, γ) with the highest cross-validation accuracy is selected as the model parameters. If multiple pairs of parameters have the same accuracy, the combination with the smaller penalty coefficient is selected to avoid overfitting. The SVM model is retrained on the entire training set using the selected pair (C, γ) to obtain the classification hyperplane equation. The model performance is evaluated using a reserved test set, and classification accuracy, precision, recall, and F1 score are calculated. If the classification accuracy on the test set is lower than a preset threshold, such as 85%, the search range is adjusted or more training data is introduced for retraining. The model that passes the final validation is the state classification model, which can output the corresponding stable state category for the input standard delay sequence. During operation, the system can periodically collect new labeled data to perform incremental training or periodic retraining of the model to adapt to the drift of delay characteristics caused by environmental changes.

[0094] The standard delayed sequence is input into the trained state classification model, and the model outputs the current stable state of the region. The stable state includes three categories: steady-state response, damped oscillation, and divergent instability. The steady-state response indicates that the control system is operating smoothly, the damped oscillation indicates that the system has periodic fluctuations but tends to be stable, and the divergent instability indicates that the system response has the risk of getting out of control.

[0095] Based on the steady-state condition and regional variation distribution, a linear weighted sum is performed using preset weight coefficients to obtain the regional ranking weights. The preset weight coefficients are determined using a multi-objective optimization method. During system initialization, with the objectives of maximizing purification efficiency and minimizing energy consumption, a particle swarm optimization algorithm is used to search within the preset weight coefficient range to select the optimal combination of weight coefficients that achieves the comprehensive objective. The steady-state condition is converted into a quantified score: steady-state response is assigned a first preset value, damped oscillation is assigned a second preset value, and divergent instability is assigned a third preset value, with the first preset value being greater than the second, which in turn is greater than the third. The concentration gradient variance in the regional variation distribution is normalized and mapped to the same numerical interval as the steady-state score to obtain the pollution urgency score. The steady-state score and the pollution urgency score are multiplied by their respective preset weight coefficients and then summed to obtain the regional ranking weights. The regional ranking weights are sorted in descending order; regions with larger weight values ​​indicate higher pollution removal urgency and stronger control system execution capabilities, and should be prioritized for wind speed adjustment strategies, ultimately resulting in a regional priority ranking.

[0096] In step S16, based on the area priority sorting, the airflow field characteristics, and the real-time wind speed, a wind speed consistency check is performed to obtain an accuracy coefficient, and a configuration association matching is performed to obtain a ventilation configuration, including:

[0097] Based on the region priority sorting, the concentration difference rate between each pair of sensor nodes in each segmented region is calculated, and sensor nodes corresponding to concentration difference rates not greater than a preset concentration difference rate threshold are selected to obtain consistent nodes.

[0098] The accuracy coefficient is obtained by calculating the difference between the real-time wind speed of the consistency node and the preset expected wind speed.

[0099] When the accuracy coefficient is lower than the preset accuracy threshold, the preset reference sampling frequency is multiplied by the preset adjustment coefficient to obtain the updated sampling frequency;

[0100] When the accuracy coefficient is not lower than the accuracy threshold, the reference sampling frequency is directly used as the update sampling frequency;

[0101] Based on the update sampling frequency and the airflow field characteristics, the ventilation configuration is obtained by configuration association matching through a preset strategy library.

[0102] Specifically, based on region priority, each segmented region is processed sequentially from highest to lowest priority. For the currently processed segmented region, the real-time concentration values ​​of all sensor nodes within that region are obtained. For each pair of sensor nodes within the region, the absolute value of the difference between the real-time concentration values ​​of the two nodes is calculated, and then divided by the average of the real-time concentration values ​​of the two nodes to obtain the concentration difference rate of that pair of nodes. The concentration difference rate is compared with a preset concentration difference rate threshold, which is determined through historical data statistics. Concentration difference rate data of all sensor node pairs are collected during a preset period of continuous operation under normal system conditions. The percentiles of these data are calculated, and the 80th percentile value is taken as the concentration difference rate threshold. When the concentration difference rate is not greater than this threshold, it indicates that the concentration data of the two nodes are consistent. Sensor nodes that meet the condition of a concentration difference rate not greater than the threshold when compared with all other nodes in the region are selected as consistent nodes, and the data of these nodes are considered reliable.

[0103] The system obtains the real-time wind speed corresponding to the consistency node. The real-time wind speed is the current air supply speed value of the air supply terminal device in the area, fed back by the ventilation system controller. It then obtains the preset expected wind speed from the system configuration. The expected wind speed is the target air supply speed value calculated for the area based on the pollution load. The difference between the preset expected wind speed and the real-time wind speed is calculated, and its absolute value is used to obtain the accuracy coefficient. The expected wind speed is obtained by obtaining the arithmetic mean of the real-time concentration values ​​of all consistency nodes in the area to obtain the regional average concentration. Next, a preset regional target concentration threshold is obtained, which is set according to the limits of corresponding pollutants in the national indoor air quality standards. Then, the difference between the regional average concentration and the regional target concentration threshold is calculated to obtain the concentration deviation value. If the concentration deviation value is greater than zero, it indicates that the current concentration exceeds the standard, and the wind speed needs to be increased. If the concentration deviation value is less than or equal to zero, it indicates that the concentration meets the standard, and the wind speed can be reduced to a maintenance level. Finally, the absolute value of the concentration deviation value is multiplied by a dimensionless proportionality coefficient, and then multiplied by the base wind speed value to obtain the wind speed adjustment amount. The base wind speed value and the wind speed adjustment amount are added together to obtain the expected wind speed. The base wind speed value is the arithmetic mean of the real-time wind speeds of all consistent nodes in the region; the proportional coefficient is determined by simulation or experimental calibration based on the historical operating data of the region to optimize the concentration decay rate under the same concentration deviation conditions, and its value ranges from 0.1 to 1.0; the proportional coefficient is selected as the preset value by applying a set of step disturbances with known concentration deviations to the region during the system debugging phase, testing the concentration decay time under different proportional coefficients, and selecting the proportional coefficient that minimizes the decay time.

[0104] The accuracy coefficient is compared with a preset accuracy threshold, which is determined through historical data statistics. Accuracy coefficient data for all control commands collected during a preset period of continuous operation under normal system conditions are collected, and the percentile of these data is calculated. The 80th percentile value is taken as the accuracy threshold. When the accuracy coefficient is lower than the preset accuracy threshold, it indicates that the current control command execution accuracy is high, and the sampling frequency needs to be increased to more sensitively capture environmental changes. The preset reference sampling frequency is multiplied by a preset adjustment coefficient to obtain the update sampling frequency. The reference sampling frequency is set according to the maximum sampling capability of the sensor hardware. The adjustment coefficient is determined by conducting multiple sets of comparative experiments during the system initialization phase, testing the system response effect under different adjustment coefficients, and selecting the adjustment coefficient that optimizes the system response efficiency as the preset value. When the accuracy coefficient is not lower than the preset accuracy threshold, it indicates that the current control command execution accuracy is insufficient or at a normal level. No adjustment of the sampling frequency is needed; the reference sampling frequency is directly used as the update sampling frequency.

[0105] Based on the updated sampling frequency and airflow field characteristics, a pre-defined strategy library is used for configuration matching to obtain the ventilation configuration. The strategy library is pre-built during system initialization, employing a mapping table structure. It takes the combination of airflow field characteristics and the updated sampling frequency as input conditions and the corresponding ventilation configuration scheme as the output result. Airflow field characteristics are categorized into four types: laminar flow, vortex flow, jet flow, and turbulent flow. The updated sampling frequency is divided into three levels: high frequency, medium frequency, and low frequency, resulting in twelve combinations. The ventilation configuration scheme corresponding to each combination is determined through experimental calibration. During system debugging, for each combination of airflow field characteristics and sampling frequency, multiple sets of comparative experiments are conducted to test the purification effect under different air supply modes, wind speed setpoints, and pulse parameters. The set of parameters with the optimal purification efficiency is selected as the ventilation configuration scheme for that combination. The ventilation configuration scheme includes operating parameters such as air supply mode, wind speed setpoint, pulse frequency, and damper opening. During step execution, the updated sampling frequency and airflow field characteristics of the current processing area are used as query conditions to match the corresponding ventilation configuration scheme in the strategy library, outputting the ventilation configuration for that area to guide the adjustment of the air supply wind speed. Process each segmented region sequentially according to its regional priority until all regions have generated corresponding ventilation configurations.

[0106] In step S17, based on the ventilation configuration and the external environmental load, hysteresis parameter optimization is performed to obtain an updated parameter set, and performance verification is performed to obtain the degree of efficiency improvement, including:

[0107] According to the ventilation configuration, control commands are issued to the actuator, the timestamp of the command issuance and the timestamp of the action feedback are recorded, and the difference between the timestamp of the command issuance and the timestamp of the action feedback is calculated to obtain the response delay sequence;

[0108] Based on the response delay sequence and the external environmental load, a curve fitting is performed using a local weighted regression algorithm to obtain an adaptive curve, and the interval in the adaptive curve where the response delay exceeds the preset response delay range is marked to obtain the hysteresis interval.

[0109] Based on the hysteresis interval, the acceleration threshold of the servo motor is increased to a preset high acceleration threshold to obtain an updated parameter set;

[0110] Based on the updated parameter set, control commands are issued to the actuator to obtain the controlled variable data stream, and the time difference between the controlled variable data stream deviating from the preset target range and recovering to the target range is calculated to obtain the current convergence time;

[0111] The difference between the pre-stored historical average time and the current convergence time is calculated and divided by the historical average time to obtain the degree of efficiency improvement.

[0112] Specifically, based on the ventilation configuration, the control system issues control commands to the actuators. The control system converts the ventilation configuration into control signals that the actuators can recognize. Simultaneously with issuing the command, the control system records the current time as a command issuance timestamp. Upon receiving the command, the actuator begins its action and, upon completion, sends a completion signal back to the control system. The control system records the moment the feedback signal is received as an action feedback timestamp. For multiple consecutively issued control commands, the above process is repeated to obtain a sequence of command issuance timestamps and an action feedback timestamp sequence. The response delay for each command is obtained by subtracting the command issuance timestamp from the action feedback timestamp. All response delays are then arranged in chronological order to obtain a response delay sequence.

[0113] Based on the response delay sequence and external environmental load, curve fitting is performed using a locally weighted regression algorithm. The algorithm uses external environmental load as the independent variable and response delay as the dependent variable. For each data point, a subset of data within a preset bandwidth is selected. The bandwidth size is determined through cross-validation, and the bandwidth value that minimizes the fitting error is chosen. For each data point, a weight is calculated based on its distance from other data points within the bandwidth; the closer the distance, the greater the weight, and the farther the distance, the smaller the weight. A cubic kernel function is used for weight calculation, meaning the weight is inversely proportional to the cube of the distance. Within the bandwidth, a straight line is found that minimizes the sum of squared weighted residuals of all data points within the bandwidth, yielding the locally fitted value at that data point. After traversing all data points, the locally fitted values ​​are connected to obtain the fitness curve.

[0114] The adaptive curve is compared with a preset response delay range, which consists of a lower threshold and an upper threshold. These thresholds are determined through historical data statistics. All response delay data are collected during a preset period of continuous operation under normal system conditions. The percentiles of these data are calculated, and the second percentile is used as the lower threshold, and the ninety-eighth percentile as the upper threshold. Intervals in the adaptive curve where the response delay exceeds this range are identified and marked as hysteresis intervals. Hysteresis intervals indicate that, under the corresponding external environmental load conditions, the actuator's response capability is insufficient or overly sensitive, posing a risk of control lag.

[0115] Based on the lag interval, the acceleration threshold of the servo motor is increased from its current value to a preset high acceleration threshold. The acceleration threshold is the maximum allowable acceleration upper limit among the servo motor control parameters. The high acceleration threshold is determined through experimental calibration. During the system debugging phase, multiple sets of comparative experiments are conducted for the external environmental load conditions corresponding to the lag interval, testing the system response effect under different acceleration thresholds. The acceleration threshold that reduces the response delay to the preset range without causing oscillation is selected as the high acceleration threshold. The increased acceleration threshold is then combined with other unadjusted control parameters to form an updated parameter set.

[0116] It should be noted that the initial acceleration threshold of the servo motor is 60% of the nominal maximum acceleration indicated on the motor nameplate; if the nameplate does not indicate this, it is determined experimentally by issuing a step speed command to the servo motor and gradually increasing the acceleration until the motor loses steps or an overcurrent alarm occurs, and 50% of this critical value is taken as the initial acceleration threshold.

[0117] Based on the updated parameter set, control commands are issued to the actuators. During command execution, the controlled variable data stream is continuously collected. The controlled variable is indoor static pressure or pollutant concentration, which is collected in real time by corresponding sensors to form time series data. The preset target range is the expected value range of the controlled variable, which is set according to process requirements. The time difference for the controlled variable data stream to recover to the target range after deviating from the preset target range is calculated. Specifically, the controlled variable value is monitored, and when the controlled variable value first leaves the target range, this moment is recorded as the deviation moment; the controlled variable value is monitored again, and when the controlled variable value first enters the target range and remains within this range for a preset stable duration, this moment is recorded as the recovery moment; the recovery moment is subtracted from the deviation moment to obtain the current convergence time. Convergence time reflects the time required for the system to recover from a disturbance to stability. The preset stability time is determined based on the noise level of the controlled variable. During the system initialization phase, data of the controlled variable running continuously for 60 seconds under steady-state conditions is collected, and its standard deviation is calculated. Three times the standard deviation is taken as the allowable fluctuation band. When the controlled variable is continuously within the target range and the fluctuation amplitude does not exceed the allowable fluctuation band for 2 seconds, it is determined to be stable. This 2 seconds is the preset stability time.

[0118] The system retrieves the pre-stored historical average time from its storage. This historical average time is a continuously updated statistic during system operation. After each performance check, the current convergence time is added to the historical dataset, and the arithmetic mean of the historical dataset is recalculated as the updated historical average time. The difference between the pre-stored historical average time and the current convergence time is calculated, and its absolute value is taken to obtain the convergence time reduction. Dividing the convergence time reduction by the historical average time yields the efficiency improvement. The efficiency improvement is expressed as a decimal or percentage.

[0119] In step S18, based on the degree of efficiency improvement and the concentration time series, deviation driving parameters are adjusted to generate optimized control commands, and performance verification and solidification are performed to obtain a continuous improvement path, including:

[0120] Align the degree of efficiency improvement with the concentration time series by timestamp to construct an efficiency-concentration matrix, and extract the dynamic change features of concentration using principal component analysis algorithm based on the efficiency-concentration matrix to obtain the concentration evolution trajectory;

[0121] The deviation between the concentration evolution trajectory and the preset baseline concentration trajectory is calculated and multiplied by the efficiency improvement degree to obtain a correction deviation vector. Based on the correction deviation vector, the adjustment amount is calculated through a pre-constructed linear mapping function to obtain the control adjustment parameters.

[0122] Based on the control adjustment parameters, an optimized control command is generated, and based on the optimized control command, a control command is issued to the actuator to obtain the pollutant decay rate;

[0123] When the pollutant decay rate is not less than the preset improvement standard, the optimization control command is directly used as the continuous improvement path;

[0124] When the pollutant decay rate is less than the improvement standard, the concentration time series is updated, and the optimized control command is regenerated until the pollutant decay rate is not less than the preset improvement standard, thus obtaining a continuous improvement path.

[0125] Specifically, the efficiency improvement is aligned with the concentration time series by timestamp to construct an efficiency-concentration matrix. The alignment operation uses the timestamp as an index, taking the efficiency improvement as a global baseline value and associating it with the concentration values ​​of all sensor nodes at each sampling time, forming a two-dimensional matrix. The rows of the matrix correspond to the sampling time, the columns correspond to the concentration values ​​of each sensor node at that time, and an additional column at the end of the matrix records the efficiency improvement value.

[0126] Based on the efficiency-concentration matrix, principal component analysis (PCA) is used to extract dynamic concentration change features. The specific process of PCA is as follows: for each concentration value in the matrix, subtract the arithmetic mean of the concentration values ​​at all times for that sensor node to obtain a centered matrix. The centered matrix is ​​then transposed, multiplied by itself, and divided by the sample size minus one. The eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvalues ​​are sorted from largest to smallest, and the eigenvector with the largest eigenvalue is selected as the first principal component. The centered matrix is ​​multiplied by the eigenvector of the first principal component to obtain the first principal component score vector. This score vector reflects the main trend of particle concentration changes throughout the monitoring area and is used as the concentration evolution trajectory.

[0127] The deviation between the concentration evolution trajectory and the preset baseline concentration trajectory is calculated and multiplied by the efficiency improvement level to obtain the correction deviation vector. The preset baseline concentration trajectory is a time series curve reflecting the expected concentration change pattern. This curve is constructed during the system initialization phase using historical data statistics, taking the 70th percentile of the concentration data at each moment under historical normal operating conditions to form the baseline concentration trajectory curve. The current value of the concentration evolution trajectory reflects the current pollution level, and its first-order difference reflects the direction and rate of concentration change. The current value of the concentration evolution trajectory is compared with the value of the baseline concentration trajectory at the current moment, and the difference is calculated to obtain the concentration deviation value. The first-order difference value of the concentration evolution trajectory is compared with the first-order difference value of the baseline concentration trajectory at the current moment, and the difference is calculated to obtain the rate of change deviation value. The concentration deviation value and the rate of change deviation value are multiplied by the efficiency improvement level respectively to obtain the weighted corrected deviation value. The two weighted corrected deviation values ​​together constitute the correction deviation vector.

[0128] Based on the corrected deviation vector, the adjustment amounts of each control parameter are calculated using a pre-constructed linear mapping function. The linear mapping function is constructed as follows: during system debugging, different combinations of concentration deviation values ​​and rate of change deviation values ​​are set in a laboratory environment or simulation platform. For each combination, different control parameter adjustment amounts are tested, and the adjustment amounts that optimize purification efficiency are recorded, forming a dataset of the correspondence between the deviation vector and the control parameter adjustment amounts. A mapping coefficient matrix is ​​obtained by fitting the dataset using a multiple linear regression method. The linear mapping function is obtained by multiplying the corrected deviation vector by the mapping coefficient matrix, resulting in a control parameter adjustment vector. This vector includes specific parameters such as wind speed setpoint adjustment, damper opening adjustment, fan frequency adjustment, and air supply mode switching indicator. The current control parameters are obtained from the current ventilation configuration, and the corresponding control parameter adjustment amounts are added to obtain the optimized control command.

[0129] According to the optimized control command, control commands are issued to the actuator. During the execution of the command, the changes in pollutant concentration are continuously monitored. The pollutant concentration is collected in real time by a particle concentration sensor to form a concentration time series. Starting from the moment the command is issued, the concentration value is recorded once at a preset sampling interval for a preset duration. The concentration value at the beginning time is subtracted from the concentration value at the end time to obtain the concentration decrease, which is then divided by the monitoring duration to obtain the pollutant decay rate. The monitoring duration is the same as the monitoring duration in step S15. The preset sampling interval is the reciprocal of the sampling frequency in step S11. When the sampling frequency is once per second, the sampling interval is 1 second. In step S18, when continuously recording the pollutant concentration, a record is made once every 1 second.

[0130] The pollutant decay rate is compared with a preset improvement standard. The improvement standard threshold is determined through historical data statistics. Data on the pollutant decay rate at which the system achieves the ideal purification effect is collected during a preset period of continuous operation under normal conditions. The percentile of these data is calculated, and the 50th percentile value is taken as the improvement standard threshold. When the pollutant decay rate is not less than the improvement standard threshold, it indicates that the current optimized control command has achieved the expected purification effect, and the optimized control command is directly used as the continuous improvement path.

[0131] When the pollutant decay rate is less than the improvement standard threshold, it indicates that the current optimized control command's purification effect is insufficient, requiring iterative optimization. The newly collected concentration data during this execution is incorporated into the existing concentration time series to update it. The process of constructing the efficiency-concentration matrix, principal component analysis, extracting the concentration evolution trajectory, and generating optimized control commands is then repeated. This cycle is repeated, with each iteration updating the concentration time series based on the latest execution results, gradually approaching the improvement standard until the pollutant decay rate reaches the improvement standard threshold, thus obtaining a continuous improvement path.

[0132] Reference Figure 2 The second embodiment of the present invention provides an adaptive control system for air supply velocity based on particle concentration, comprising:

[0133] The data acquisition module is used to acquire concentration time series, particle position coordinates, environmental interference factors, real-time wind speed, airflow field characteristics, and external environmental loads;

[0134] The feature analysis module is used to perform heterogeneous feature analysis by fitting node evolution trajectories based on the concentration time series and the particle position coordinates to obtain heterogeneous distribution features;

[0135] The curve fitting module is used to compare the fluctuation amplitude based on the heterogeneous distribution characteristics and the concentration time series to obtain the peak period, and to fit the concentration trend curve based on the peak period to obtain the concentration trend curve.

[0136] The region division module is used to correct the unit concentration based on the concentration trend curve and the environmental interference factor to obtain corrected grid units, and to divide the difference regions based on the concentration difference comparison of the corrected grid units to obtain the regional difference distribution.

[0137] The weight calculation module is used to classify states based on the regional difference distribution and delay judgment to obtain stable states, and to calculate and sort regional weights based on the stable states to obtain regional priority ranking.

[0138] The configuration matching module is used to perform wind speed consistency verification based on the area priority sorting, the airflow field characteristics and the real-time wind speed, obtain the accuracy coefficient, and perform configuration association matching to obtain the ventilation configuration;

[0139] The performance verification module is used to optimize the hysteresis parameters based on the ventilation configuration and the external environmental load, obtain an updated parameter set, and perform performance verification to obtain the degree of efficiency improvement.

[0140] The output module is used to adjust the deviation driving parameters according to the degree of efficiency improvement and the concentration time series, generate optimized control instructions, and perform performance verification and solidification to obtain a continuous improvement path.

[0141] It should be noted that the particle concentration-based adaptive airflow speed control system provided in this embodiment of the invention is used to execute all the process steps of the particle concentration-based adaptive airflow speed control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0142] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0143] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for adaptive control of air supply velocity based on particle concentration, characterized in that, include: Acquire concentration time series, particle position coordinates, environmental disturbance factors, real-time wind speed, airflow field characteristics, and external environmental load; Based on the concentration time series and the particle position coordinates, heterogeneous feature analysis is performed by fitting node evolution trajectories to obtain heterogeneous distribution characteristics. Based on the heterogeneous distribution characteristics and the concentration time series, the fluctuation amplitude is compared to obtain the peak period, and the concentration trend curve is fitted based on the peak period to obtain the concentration trend curve; Based on the concentration trend curve and the environmental disturbance factor, the unit concentration is corrected to obtain the corrected grid unit. Based on the corrected grid unit, the difference region is divided according to the concentration difference comparison to obtain the regional difference distribution. Based on the regional difference distribution, state classification is performed based on delay judgment to obtain a stable state, and regional weights are calculated and sorted according to the stable state to obtain a regional priority ranking. Based on the regional priority ranking, the airflow field characteristics, and the real-time wind speed, a wind speed consistency check is performed to obtain an accuracy coefficient, and a configuration association matching is performed to obtain a ventilation configuration. Based on the ventilation configuration and the external environmental load, hysteresis parameters are optimized to obtain an updated parameter set, and performance is verified to determine the degree of efficiency improvement. Based on the degree of efficiency improvement and the concentration time series, deviation-driven parameters are adjusted to generate optimized control commands, and performance is verified and solidified to obtain a continuous improvement path.

2. The adaptive control method for air supply velocity based on particle concentration according to claim 1, characterized in that, The heterogeneous distribution characteristics are obtained by performing heterogeneous feature analysis based on the concentration time series and the particle position coordinates through node evolution trajectory fitting, including: Based on the concentration time series, the particle position coordinates, and the preset sampling frequency, the concentration change rate is obtained by calculating the particle concentration difference at adjacent sampling times for the same node. Based on the concentration change rate, the diffusion gradient distribution is obtained by calculating the ratio of the rate difference between adjacent nodes to the spatial distance. Calculate the difference between the diffusion gradient of each node and the diffusion gradient of its neighboring nodes in the diffusion gradient distribution to obtain the gradient difference, and filter out the nodes corresponding to the gradient difference exceeding a preset jump threshold to obtain local extreme points. Based on the local extreme points, cross-time node association is performed using the K nearest neighbor algorithm to obtain a sequence of common-origin nodes. Based on the sequence of common-origin nodes, curve fitting is performed using cubic spline interpolation to obtain a continuous evolution trajectory. The starting position, direction of movement, and spatial coverage of the continuous evolution trajectory are extracted to obtain the heterogeneous distribution characteristics.

3. The adaptive control method for air supply velocity based on particle concentration according to claim 1, characterized in that, The step of comparing fluctuation amplitudes based on the heterogeneous distribution characteristics and the concentration time series to obtain peak periods, and fitting a concentration trend curve based on the peak periods to obtain a concentration trend curve, includes: Based on the heterogeneous distribution characteristics, the concentration time series is filtered and denoised using a weighted moving average algorithm to obtain a smoothed series. The difference between local maxima and local minima in the smoothed sequence is calculated to obtain the fluctuation amplitude; The peak time period is obtained by filtering out the time segments in the smoothed sequence where the fluctuation amplitude exceeds a preset fluctuation amplitude threshold; Polynomial fitting was performed on the concentration values ​​during the peak period to obtain the concentration trend curve.

4. The adaptive control method for air supply velocity based on particle concentration according to claim 1, characterized in that, The process of correcting unit concentrations based on the concentration trend curve and the environmental disturbance factor to obtain corrected grid cells, and then dividing the region into different areas based on concentration difference comparisons using the corrected grid cells to obtain the regional difference distribution, includes: Based on the concentration time series and the preset three-dimensional spatial coordinates, the concentration grid cells are obtained by dividing the grid using a uniform spatial grid division method. When the slope of the tangent of the concentration trend curve exceeds the preset concentration change rate threshold, the interference weight is obtained by performing a weight lookup through a preset weight mapping table according to the environmental interference factor, and the concentration in the concentration grid cell is weighted and corrected according to the interference weight to obtain the corrected grid cell. When the slope of the tangent line of the concentration trend curve does not exceed the concentration change rate threshold, the concentration grid cell is directly used as the correction grid cell; Calculate the concentration difference between adjacent modified grid cells to obtain the concentration difference; When the concentration difference is greater than the preset tolerance threshold, the current corrected grid cell is directly marked as an independent region to obtain the segmented region; When the concentration difference is not greater than the tolerance threshold, the adjacent corrected grid cells are merged into the region where the current corrected grid cell is located, thus obtaining the segmented region; Calculate the variance of the concentration gradient between each segmented region to obtain the regional difference distribution.

5. The adaptive control method for air supply velocity based on particle concentration according to claim 4, characterized in that, The step of classifying states based on the regional difference distribution and delay judgment to obtain stable states, and then calculating and ranking regional weights based on the stable states to obtain regional priority ranking, includes: Based on the regional difference distribution, the real-time refresh frequency and feedback delay of each segmented region are extracted to obtain a refresh frequency sequence and a feedback delay sequence. When the variance of the refresh frequency sequence exceeds a preset frequency variance threshold, the feedback delay sequence is smoothed by a median filtering algorithm to obtain a standard delay sequence. When the variance of the refresh frequency sequence does not exceed the frequency variance threshold, the feedback delay sequence is directly used as the standard delay sequence. Based on the standard delay sequence, a stable state is obtained by classifying and identifying it using a pre-built state classification model. Based on the stable state and the regional difference distribution, a linear weighted sum is performed using preset weight coefficients to obtain the regional ranking weights. The regional ranking weights are then sorted in descending order to obtain the regional priority ranking.

6. The adaptive control method for air supply velocity based on particle concentration according to claim 5, characterized in that, The step of performing wind speed consistency verification based on the region priority sorting, the airflow field characteristics, and the real-time wind speed to obtain an accuracy coefficient, and then performing configuration association matching to obtain the ventilation configuration, includes: Based on the region priority sorting, the concentration difference rate between each pair of sensor nodes in each segmented region is calculated, and sensor nodes corresponding to concentration difference rates not greater than a preset concentration difference rate threshold are selected to obtain consistent nodes. The accuracy coefficient is obtained by calculating the difference between the real-time wind speed of the consistency node and the preset expected wind speed. When the accuracy coefficient is lower than the preset accuracy threshold, the preset reference sampling frequency is multiplied by the preset adjustment coefficient to obtain the updated sampling frequency; When the accuracy coefficient is not lower than the accuracy threshold, the reference sampling frequency is directly used as the update sampling frequency; Based on the update sampling frequency and the airflow field characteristics, the ventilation configuration is obtained by configuration association matching through a preset strategy library.

7. The adaptive control method for air supply velocity based on particle concentration according to claim 1, characterized in that, The process of optimizing hysteresis parameters based on the ventilation configuration and the external environmental load to obtain an updated parameter set, and then performing performance verification to determine the degree of efficiency improvement, includes: According to the ventilation configuration, control commands are issued to the actuator, the timestamp of the command issuance and the timestamp of the action feedback are recorded, and the difference between the timestamp of the command issuance and the timestamp of the action feedback is calculated to obtain the response delay sequence; Based on the response delay sequence and the external environmental load, a curve fitting is performed using a local weighted regression algorithm to obtain an adaptive curve, and the interval in the adaptive curve where the response delay exceeds the preset response delay range is marked to obtain the hysteresis interval. Based on the hysteresis interval, the acceleration threshold of the servo motor is increased to a preset high acceleration threshold to obtain an updated parameter set; Based on the updated parameter set, control commands are issued to the actuator to obtain the controlled variable data stream, and the time difference between the controlled variable data stream deviating from the preset target range and recovering to the target range is calculated to obtain the current convergence time; The difference between the pre-stored historical average time and the current convergence time is calculated and divided by the historical average time to obtain the degree of efficiency improvement.

8. The adaptive control method for air supply velocity based on particle concentration according to claim 1, characterized in that, The process involves adjusting deviation-driven parameters based on the efficiency improvement and the concentration time series, generating optimized control commands, and verifying and solidifying the performance to obtain a continuous improvement path, including: Align the degree of efficiency improvement with the concentration time series by timestamp to construct an efficiency-concentration matrix, and extract the dynamic change features of concentration using principal component analysis algorithm based on the efficiency-concentration matrix to obtain the concentration evolution trajectory; The deviation between the concentration evolution trajectory and the preset baseline concentration trajectory is calculated and multiplied by the efficiency improvement degree to obtain a correction deviation vector. Based on the correction deviation vector, the adjustment amount is calculated through a pre-constructed linear mapping function to obtain the control adjustment parameters. Based on the control adjustment parameters, an optimized control command is generated, and based on the optimized control command, a control command is issued to the actuator to obtain the pollutant decay rate; When the pollutant decay rate is not less than the preset improvement standard, the optimization control command is directly used as the continuous improvement path; When the pollutant decay rate is less than the improvement standard, the concentration time series is updated, and the optimized control command is regenerated until the pollutant decay rate is not less than the preset improvement standard, thus obtaining a continuous improvement path.

9. A system for adaptive control of air supply velocity based on particle concentration, characterized in that, include: The data acquisition module is used to acquire concentration time series, particle position coordinates, environmental interference factors, real-time wind speed, airflow field characteristics, and external environmental loads; The feature analysis module is used to perform heterogeneous feature analysis by fitting node evolution trajectories based on the concentration time series and the particle position coordinates to obtain heterogeneous distribution features; The curve fitting module is used to compare the fluctuation amplitude based on the heterogeneous distribution characteristics and the concentration time series to obtain the peak period, and to fit the concentration trend curve based on the peak period to obtain the concentration trend curve. The region division module is used to correct the unit concentration based on the concentration trend curve and the environmental interference factor to obtain corrected grid units, and to divide the difference regions based on the concentration difference comparison of the corrected grid units to obtain the regional difference distribution. The weight calculation module is used to classify states based on the regional difference distribution and delay judgment to obtain stable states, and to calculate and sort regional weights based on the stable states to obtain regional priority ranking. The configuration matching module is used to perform wind speed consistency verification based on the area priority sorting, the airflow field characteristics and the real-time wind speed, obtain the accuracy coefficient, and perform configuration association matching to obtain the ventilation configuration; The performance verification module is used to optimize the hysteresis parameters based on the ventilation configuration and the external environmental load, obtain an updated parameter set, and perform performance verification to obtain the degree of efficiency improvement. The output module is used to adjust the deviation driving parameters according to the efficiency improvement degree and the concentration time series, generate optimized control instructions, and perform performance verification and solidification to obtain a continuous improvement path.

Images