Intelligent partition purification control system for experimental environment

Abstract

The invention relates to the technical field of air purification and intelligent control, particularly discloses an intelligent partition purification control system for an experimental environment, and aims to solve the problems that a partition strategy is rigid, pollution situation response is insufficient, and multi-target collaborative optimization is difficult. The system comprises an environment situation awareness module, a dynamic partition decision module, a flexible execution scheduling module and a global collaborative optimization module. A dynamic pollution situation field is constructed in real time, a virtual purification partition is solved, fixed and mobile purification units are dynamically scheduled to execute a collaborative purification task, decision parameters are calibrated on line based on a multi-target optimization model, and global dynamic optimization of purification efficiency, response speed and operation energy consumption is achieved. And the accurate and efficient control capability of experimental environment pollution is improved.

Description

Technical Field

[0001] This invention relates to the field of air purification and intelligent control technology, specifically to an intelligent zoned purification control system for experimental environments. Background Technology

[0002] With increasingly stringent requirements for the precision control of key parameters such as cleanliness, temperature, humidity, and pollutant concentration in experimental environments in fields such as biomedicine, precision manufacturing, and cutting-edge scientific research, the level of intelligence in environmental control systems has become a core element in ensuring the reliability of experimental data, product quality, and personnel safety. Intelligent control of experimental environments aims to achieve precise response and efficient management of complex and dynamic pollution scenarios by integrating sensing, decision-making, and execution units.

[0003] Among these, intelligent zoned purification control of the experimental environment is a key technological direction for improving environmental safety assurance capabilities. This technology divides the experimental space into different functional or cleanliness zones and implements differentiated airflow organization and purification strategies to effectively isolate pollution sources, prevent cross-contamination, and optimize energy consumption. Its core objective is to build a highly efficient system capable of adapting to environmental changes, dynamically adjusting control boundaries, and coordinating multiple execution units.

[0004] Existing technologies largely rely on pre-defined fixed physical zones and control logic based on simple threshold triggers. These systems lack real-time, precise sensing capabilities for the dynamic diffusion of pollution sources, personnel activity trajectories, and airflow coupling effects between multiple zones. Traditional solutions struggle to establish closed-loop optimization models between multi-dimensional environmental data and multi-dimensional control parameters of purification equipment, resulting in lag in control response, blind spots in purification coverage, and low energy efficiency. While some improved solutions introduce mobile purification equipment or rule-based state matching, their zoning strategies remain rigid, unable to dynamically reconstruct the topology based on the real-time evolution of pollution events. The movement paths of mobile devices are limited and lack coordination, leading to global control strategies often getting bogged down in local optimization. This fails to achieve overall optimization of purification efficiency, response speed, and operating energy consumption under multiple constraints. Therefore, achieving real-time, accurate sensing of pollution conditions within the experimental environment, dynamically generating optimal purification zones, and collaboratively scheduling flexible execution resources has become a pressing technical challenge in this field. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent zoned purification control system for experimental environments, in order to solve the technical contradictions in the prior art, such as rigid zoning strategies, insufficient perception and response capabilities to pollution situations, low efficiency of multi-execution unit collaboration, and difficulty in achieving global optimal control under multiple constraints such as purification efficiency, response speed, and operating energy consumption.

[0006] To achieve the above objectives, this invention provides an intelligent zoned purification control system for experimental environments. The system includes: The environmental situation awareness module is used to collect and fuse multi-dimensional state data in the experimental environment in real time to construct a dynamic pollution situation field. The dynamic zoning decision module is used to calculate and generate a virtual purification zone topology in real time based on the dynamic pollution situation field. The flexible execution scheduling module is used to assign tasks and plan action paths to the purification execution units according to the topology of the virtual purification partition. The global collaborative optimization module is used to establish and solve an optimization model with purification efficiency, response speed and system energy consumption as comprehensive objectives, and to perform online calibration and collaborative optimization of the operating parameters of the dynamic partition decision module and the flexible execution scheduling module. The environmental situation awareness module includes a distributed sensor network, a personnel and equipment activity tracking unit, and a data fusion and field construction unit. The distributed sensing network consists of multiple types of sensor nodes deployed in the experimental environment. The sensor types include at least laser particle counters, multi-gas component detectors, temperature and humidity sensors, and micro-differential pressure sensors. Each sensor node synchronously collects pollutant concentration, gas composition, temperature, humidity, and air pressure data at its location at a sampling frequency of no less than once per second. The personnel and equipment activity tracking unit tracks the location coordinates and movement trajectory of experimental personnel, mobile devices, and materials in real time through ultra-wideband positioning base stations and tags deployed in the environment. The data fusion and field construction unit receives raw data streams from the distributed sensor network and the personnel and equipment activity tracking unit. It uses a multi-source data spatiotemporal registration algorithm based on Kalman filtering to unify asynchronous and heterogeneous sensor data into the same spatiotemporal coordinate system. It then uses the Kriging space interpolation algorithm to calculate the pollutant concentration distribution at locations not directly monitored throughout the space in real time, based on the registered discrete point monitoring data and combined with the three-dimensional geometric model of the experimental environment and preset airflow organization parameters. This constructs a dynamic pollutant concentration scalar field covering the entire environment. The preset airflow organization parameters include the main airflow velocity vector field estimated based on micro-differential pressure sensor data and a simplified computational fluid dynamics model. The dynamic zoning decision module includes a pollution source and risk zone identification unit, a zoning topology calculation unit, and a zoning attribute definition unit. The pollution source and risk area identification unit interfaces with the dynamic pollution concentration scalar field and personnel activity trajectory data output by the data fusion and field construction unit. By setting concentration gradient threshold and time change rate threshold, it automatically identifies areas with rapidly rising concentrations and marks them as active pollution sources. At the same time, based on the real-time location of personnel and key equipment, combined with preset safety concentration limits, it delineates risk areas that need to be protected. The partitioned topology solving unit takes the identified pollution sources and risk zones as input, uses the physical space as nodes in graph theory, and uses the Euclidean distance, connectivity, and estimated airflow influence intensity between nodes as edge weights to construct an undirected weighted graph of the environmental space. It then runs a pollution diffusion blocking path search program based on an improved Dijkstra algorithm. The improvement lies in that, in the relaxation operation of the standard Dijkstra algorithm, not only is the cumulative path weight considered, but a dynamic penalty factor is also introduced. This dynamic penalty factor is proportional to the real-time pollutant concentration at the current path endpoint node, prioritizing the exploration of blocking paths in high-concentration areas. The program starts with each active pollution source and uses all risk zones as potential endpoints, searching and marking the critical paths and boundaries in the weighted graph that can most effectively block the diffusion of pollutants from the source to the risk zone. These marked paths and boundaries are spatially connected end-to-end, forming closed virtual boundaries, thereby dynamically dividing the environment into multiple mutually isolated virtual purification zones. The partition attribute definition unit assigns control attributes to each generated virtual purification partition. The attributes include at least the current average pollutant concentration, the highest pollutant concentration, the area and volume of the partition, the identification of the equipment and personnel contained in the partition, and the partition purification priority coefficient calculated based on the pollution source intensity and risk level.

[0007] In some embodiments, when the partitioned topology solving unit runs the pollution diffusion barrier path search program, the weight of the edges in the constructed undirected weighted graph is determined by three factors: the first factor is the physical distance between nodes, the second factor is the airflow intensity from upstream to downstream nodes estimated by the data fusion and field construction unit, and the third factor is whether there are physical barriers or wind curtains between nodes; the final weight of the edge is the weighted product of these three factors, calculated using the following formula: ,in, This represents the edge weight from node i to node j. Let be the Euclidean distance from node i to node j. Let ϵ be the estimated airflow intensity from node i to node j, where ϵ is a local constant. As a barrier factor, These are the weighting coefficients for distance, airflow, and obstruction, respectively. The global collaborative optimization module dynamically adjusts the parameters.

[0008] In some embodiments, the flexible execution scheduling module includes an execution unit resource library, a task-resource matching unit, and a path planning and collaborative control unit; The execution unit resource library registers and manages all available purification execution units in the system, including fixed fresh air units, fixed high-efficiency air filters, ceiling-mounted circulating fans, and mobile air purification robots with autonomous movement capabilities. The attribute records of each execution unit in the resource library include its unique identifier, type, real-time location coordinates, rated purification air volume, effective radius of action, movement speed, current working status, and remaining energy. The task-resource matching unit receives the virtual purification partition topology and its attributes from the dynamic partitioning decision module. For each virtual purification partition, it calculates the required total purification airflow based on the partition's spatial geometry, purification priority coefficient, and the target cleanliness level. Based on the real-time status of the execution unit resource library, it employs a hybrid matching algorithm combining the Hungarian algorithm and a greedy strategy to allocate a suitable combination of purification execution units to each partition. In the cost calculation of the hybrid matching algorithm, distance cost has a higher weight than energy cost. That is, by increasing the weight coefficient of the distance term, the available unit closest to the geometric center of the partition is preferentially selected while meeting the purification airflow requirements, thereby minimizing overall start-up, shutdown, and movement energy consumption.

[0009] In some embodiments, when the task-resource matching unit assigns a mobile air purification robot to a virtual purification zone, the matching algorithm not only considers the distance between the robot's real-time position and the zone center, but also includes the estimated pollutant concentration on the path required for the robot to reach the zone as an additional cost item in the calculation, and assigns a discount coefficient to the cost of high-concentration areas on the path, wherein the discount coefficient is a positive coefficient less than 1.

[0010] In some embodiments, the path planning and collaborative control unit is specifically designed for mobile air-purifying robots. Based on the instructions issued by the task-resource matching unit, this unit plans the optimal travel path for mobile robots that need to go to a designated zone to perform tasks. The path planning adopts an algorithm that integrates dynamic window method and time elastic band. This algorithm not only considers static obstacles, but also incorporates the activity trajectories of other mobile robots and personnel in real time as dynamic obstacles for obstacle avoidance calculation. At the same time, this unit generates collaborative control instructions to coordinate the wind speed, wind direction and start-stop sequence of multiple fixed and mobile execution units in the same zone.

[0011] In some embodiments, the global collaborative optimization module includes a multi-objective optimization model, an online parameter calibrator, and a system performance evaluation and feedback unit; The multi-objective optimization model constructs a mathematical function with the goal of minimizing the overall system operating cost. This function is a weighted sum of purification efficiency cost, response time cost, and energy consumption cost. The online parameter calibrator is embedded in the key decision-making stages of the dynamic partitioning decision module and the flexible execution scheduling module. Specifically, it includes the edge weight calculation formula of the pollution diffusion blocking path search algorithm in the dynamic partitioning decision module and the preference weight of the task-resource matching algorithm in the flexible execution scheduling module. The online parameter calibrator receives historical performance data from the system performance evaluation and feedback unit at a time interval of minutes and dynamically adjusts the above-mentioned internal algorithm parameters using an online learning strategy based on gradient descent. The system performance evaluation and feedback unit continuously monitors the pollutant concentration change curves of each virtual purification zone, the actual energy consumption data of each execution unit, and the overall task completion time. It calculates the periodic purification efficiency, energy consumption per unit of purification volume, and average response delay index, and inputs these indexes as performance feedback to the online parameter calibrator.

[0012] In some embodiments, the purification efficiency cost is quantified by the sum of squares of the deviations between the actual pollutant concentration decrease rate and the target decrease rate in all virtual purification zones; the response time cost is quantified by the time delay from the identification of a pollution event to the attainment of the target concentration at key points within the zone; and the energy consumption cost is quantified by the sum of the power integrals of all fixed and mobile execution units in operation.

[0013] In some embodiments, the weight coefficients of the multi-objective optimization model are not fixed. The system presets multiple typical scenario modes, including daily monitoring mode, emergency pollution response mode, and energy-saving operation mode. In different scenario modes, the weight coefficients of purification efficiency, response speed, and energy consumption in the global cost function have different preset combinations.

[0014] In some embodiments, the data fusion and field construction unit further estimates the main airflow velocity vector field based on micro differential pressure sensor data and a simplified computational fluid dynamics model.

[0015] In some embodiments, the optimization objective of the hybrid matching algorithm is to minimize the overall start-up, shutdown, and movement energy consumption of the execution unit while meeting the air volume requirements of the partition purification, and to prioritize scheduling the available unit closest to the geometric center of the partition.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a dynamic pollution situation field through an environmental situation awareness module, and based on this, a dynamic zoning decision module calculates the virtual purification zone topology in real time, completely changing the traditional control paradigm that relies on fixed physical zones. The system can accurately perceive the generation and diffusion dynamics of pollution sources and dynamically reconstruct the purification boundary based on this, realizing a fundamental shift from "static zoning and passive response" to "dynamic zoning and active containment," significantly improving the response accuracy and purification efficiency for sudden or mobile pollution sources, and effectively eliminating control blind spots.

[0017] 2. This invention achieves unified optimized scheduling and collaborative control of fixed and mobile purification execution resources through deep collaboration between the flexible execution scheduling module and the global collaborative optimization module. The task-resource matching and path planning algorithm not only considers spatial distance and task requirements but also introduces optimization factors such as purification benefits along the way, ensuring that every scheduling of the execution unit serves to improve overall purification efficiency. The multi-objective optimization model and online parameter calibration mechanism ensure that the system can achieve dynamic balance and continuous optimization among multiple mutually constraining objectives such as purification efficiency, response speed, and operating energy consumption, overcoming the shortcomings of traditional solutions that often pursue a single indicator while neglecting overall efficiency.

[0018] 3. This invention establishes a complete closed-loop intelligent control system encompassing "perception-decision-execution-optimization." The global collaborative optimization module continuously calibrates front-end decision parameters through online learning, enabling the system to self-improve based on operational experience. This closed-loop optimization mechanism allows the system to continuously adapt to the specific airflow characteristics, usage habits, and pollution patterns of the experimental environment, maintaining a highly efficient, stable, and economical operating state over the long term, providing a reliable technical foundation for achieving intelligent management of the experimental environment. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram of the overall technical architecture of the intelligent zoned purification control system for experimental environments proposed in this invention; Figure 2 This is a schematic diagram illustrating the core principle framework of dynamic pollution situation field construction and dynamic zoning decision-making in this invention; Figure 3 This is a flowchart illustrating the task-resource matching and path planning logic of the flexible execution scheduling module in this invention. Figure 4This is a schematic diagram of the multi-level collaborative control and data flow of the fixed and mobile purification execution units in this invention; Figure 5 This is a schematic diagram of the multi-objective optimization and online parameter calibration principle framework of the global collaborative optimization module in this invention. Detailed Implementation Example

[0020] This invention provides an intelligent zoned purification control system for experimental environments. Please refer to the appendix. Figure 1 This system is a closed-loop intelligent control system integrating an environmental situation awareness module, a dynamic zoning decision-making module, a flexible execution scheduling module, and a global collaborative optimization module. Deployed in an experimental environment with a complex internal structure, this system typically includes multiple workbenches, fume hoods, instrument and equipment storage areas, personnel activity areas, and material channels. The core objective of the system is to break away from the traditional purification model that relies on fixed physical barriers for area isolation. By sensing the pollution situation within the environment in real time, it dynamically constructs virtual purification zones and intelligently schedules and collaboratively controls various purification execution units, thereby achieving globally optimal pollution control under multiple constraints such as purification efficiency, response speed, and system energy consumption.

[0021] The environmental situation awareness module is the data foundation for the system to perceive the environmental state. Please refer to the appendix. Figure 2 This module consists of a distributed sensor network, a personnel and equipment activity tracking unit, and a data fusion and field construction unit. The distributed sensor network comprises numerous sensor nodes of various types, strategically deployed according to the three-dimensional structure of the experimental environment. Sensor nodes are primarily installed on grid nodes in the ceiling, at specific heights on walls, and around critical equipment such as fume hoods, biosafety cabinets, and reaction devices. Sensor types include at least laser particle counters, multi-gas component detectors, temperature and humidity sensors, and micro-differential pressure sensors. Laser particle counters are used to monitor the number and concentration of suspended particulate matter in the air in real time, covering a particle size range from 0.3 micrometers to 10 micrometers. Multi-gas component detectors are configured according to the characteristic pollutants that may be generated in the experiment, such as monitoring the concentration of volatile organic compounds, carbon monoxide, carbon dioxide, ammonia, and hydrogen sulfide. Temperature and humidity sensors monitor the ambient air temperature and relative humidity. Micro-differential pressure sensors are deployed in pairs on either side of the boundary of areas with potential pressure difference requirements, such as between clean areas and potentially contaminated areas, to monitor the pressure difference between the areas. All sensor nodes are connected to the system network via wired or wireless industrial IoT protocols, and synchronously collect data on pollutant concentration, gas composition, temperature, humidity and air pressure at their location at a sampling frequency of no less than once per second, forming a raw data stream.

[0022] The personnel and equipment activity tracking unit achieves real-time tracking with centimeter-level accuracy through an ultra-wideband (UWB) positioning system deployed within the environment. This system consists of multiple UWB base stations fixed at known coordinates on the ceiling, and UWB tags worn by experimental personnel and attached to key mobile devices and material carts. The base stations continuously receive wireless signals emitted by the tags and calculate the real-time three-dimensional spatial coordinates of each tag using time difference of arrival (TDOA) or angle of arrival (AHE) algorithms. This unit not only records coordinates but also calculates movement speed and direction by analyzing the coordinate sequences, thereby identifying activity states such as stationary, walking, operating equipment, and handling materials. All real-time location, movement trajectory, and activity status data of personnel, mobile devices, and materials are encapsulated into a structured data stream.

[0023] The data fusion and field construction unit receives asynchronous, heterogeneous raw data streams from the two units mentioned above. Its primary task is to perform spatiotemporal registration of multi-source data. This unit maintains a global three-dimensional Cartesian coordinate system with a fixed corner of the experimental environment as the origin. The physical installation locations of all sensor nodes and the coordinates of the ultra-wideband base station have been precisely calibrated in this coordinate system. For each frame of sensor data, the data fusion and field construction unit first assigns it a precise timestamp, and then assigns it to a specific point in space based on the known location coordinates of the sensor. Since there may be slight differences in the sampling periods of different sensors, this unit adopts a prediction and update algorithm based on Kalman filtering. This algorithm is based on the system dynamics model and makes short-time predictions of state variables such as pollutant concentration and temperature at each monitoring point. When new sensor data arrives, the predicted values ​​are fused with the measured values ​​to obtain the optimal estimate of the state of that spatial point at that moment. This process effectively smooths measurement noise and unifies all data under the same spatiotemporal reference.

[0024] After completing spatiotemporal registration, the core task of the data fusion and field construction unit is to construct a dynamic scalar field of pollutant concentration with spatial continuity, covering the entire experimental environment. This unit stores a precise three-dimensional geometric model of the experimental environment, including the approximate outlines of all walls, partitions, and equipment. Using registered discrete-point monitoring data as known sample points, this unit employs the Kriging spatial interpolation algorithm for full-field estimation. The Kriging algorithm not only considers the distance between sample points and the points to be estimated but also characterizes the autocorrelation of spatial data through a variogram model. Specifically, the algorithm first calculates the semivariogram at different distance intervals based on the pollutant concentration values ​​of the known sample points, fitting a variogram curve. This curve reflects the continuity and variability of pollutant concentration changes in space. Then, for any location in the environment that is not directly monitored, the algorithm uses the concentrations of all known sample points within a certain range around it as input, performs a weighted summation based on the weighting coefficients determined by the variogram, and thus estimates the pollutant concentration at that point. This process is performed point-by-point on tens of thousands of points in the environment, ultimately generating a digital field of concentration distribution with adjustable resolution covering the entire environment. This field can visually display the accumulation areas, diffusion gradients, and concentration distribution of pollutants. Simultaneously, the data fusion and field construction unit, based on data from a deployed micro-differential pressure sensor network and combined with preset air supply and return inlet locations and airflow parameters, runs a simplified computational fluid dynamics steady-state model. This model discretizes the environmental space into a coarse grid and estimates the direction and magnitude of the average airflow velocity vector in the main areas by solving simplified mass and momentum conservation equations, thereby aiding in the construction of the airflow velocity vector field.

[0025] The dynamic zoning decision module executes the core spatial partitioning decision based on the dynamic field information provided by the environmental situation awareness module. Please refer to the appendix for further details. Figure 2This module includes a pollution source and risk zone identification unit, a partition topology calculation unit, and a partition attribute definition unit. The pollution source and risk zone identification unit analyzes the dynamic pollution concentration scalar field in real time. This unit sets two key thresholds: a concentration gradient threshold and a time rate of change threshold. The concentration gradient threshold is used to identify spatial boundaries where concentration changes drastically, and the time rate of change threshold is used to identify points where concentration rises rapidly over time. The system continuously scans the concentration field, calculating the spatial gradient of its concentration relative to neighboring points and the rate of change of the current concentration relative to the previous time period for any given spatial location. When both the spatial gradient and the time rate of change of a local area exceed the preset thresholds, that area is marked as an active pollution source. The system records the center coordinates, spatial range, initial concentration intensity, and intensity change trend of the pollution source. Simultaneously, this unit obtains real-time data from the personnel and equipment activity tracking unit, identifying the workstations of laboratory personnel, the locations of critical precision instruments and equipment in operation, and areas storing important samples as risk areas requiring key protection. The system presets corresponding safe concentration limits for each type of risk area.

[0026] The partitioned topology solution unit receives the identified set of active pollution sources and risk areas. This unit abstracts the physical space of the entire experimental environment as an undirected weighted graph in graph theory. Specifically, the three-dimensional environmental space is discretized into a grid on a horizontal plane, with the center point of each grid cell considered a node in the graph. The connections between nodes are the edges of the graph, and the rules for establishing edges are based on spatial connectivity: if two grid cells are spatially adjacent and there are no impenetrable physical barriers such as solid walls between them, then an edge is established between their corresponding nodes. Each edge is assigned a weight value, which characterizes the ease or cost of pollutant diffusion along the path represented by this edge. The weight value is determined by three factors. The first factor is the physical Euclidean distance between nodes; the greater the distance, the greater the weight. The second factor is the average airflow intensity from upstream nodes to downstream nodes, estimated by the data fusion and field construction unit; the greater the airflow intensity, the easier it is for pollutants to diffuse downwind, and the smaller the weight; conversely, the weight increases for upwind diffusion. The third factor is whether there are physical barriers or active barriers such as air curtains between nodes. If so, this factor is multiplied by a very large coefficient, causing a sharp increase in weight to simulate the barrier effect. The final weight of the edge is the weighted product of these three factors. ,in, This represents the edge weight from node i to node j. This is the estimated airflow intensity from direction i to j. It is a very small constant to prevent division by zero errors. It is a barrier factor. These are the weighting coefficients for distance, airflow, and obstruction, respectively. These weighting coefficients are dynamically adjusted by the global collaborative optimization module. Based on the constructed undirected weighted graph, the partitioned topology solution unit runs a pollution diffusion obstruction path search program based on an improved Dijkstra algorithm. The improvement lies in that, in the relaxation operation of the standard Dijkstra algorithm, not only is the path accumulation weight considered, but a dynamic penalty factor is also introduced. This dynamic penalty factor is proportional to the real-time pollutant concentration at the current path endpoint node, used to prioritize exploring obstruction paths in high-concentration areas. Specifically, the dynamic penalty factor... The calculation formula is ,in, The real-time pollutant concentration at the current path endpoint node n is given by k, which is a preset proportionality constant with a value range of [0.1, 1.0], used to adjust the exploration priority of high-concentration areas.

[0027] The program uses the node at the center of each identified active pollution source as the source point and the set of nodes covered by all risk areas as the target node set. The algorithm aims to find the shortest weighted path from the source point to any node in the target node set. Here, "shortest" means that the cost of pollutant diffusion along this path is the lowest and the probability is the highest. After the algorithm is executed, it marks all the shortest paths from pollution sources to risk areas. These marked paths spatially delineate the channels through which pollutants are most likely to diffuse. The partition topology solving unit then extracts these critical paths and the nodes connected by the edges with higher weights on the paths as critical boundaries that need to be "cut off" or "isolated". The system connects these boundary nodes end to end in space to form closed virtual boundary lines. These virtual boundary lines dynamically divide the entire experimental environment into several unconnected sub-regions, each of which constitutes a virtual purification partition. The number, shape, and location of the partitions change in real time as the pollution sources and risk areas change.

[0028] The partition attribute definition unit assigns detailed control attributes to each generated virtual cleanup partition. These attributes include: the arithmetic mean of pollutant concentrations across all grid nodes within the partition (i.e., the current average pollutant concentration); the maximum pollutant concentration across all grid nodes within the partition (i.e., the highest pollutant concentration); the partition's floor area (the total number of grids occupied by the partition multiplied by the area of ​​a single grid, then multiplied by the average floor height to obtain the partition's volume); a list of all personnel and equipment tags within the partition; and a crucial calculated attribute—the partition cleanup priority coefficient. This coefficient is calculated using a weighted formula based on factors such as the partition's pollution source intensity, the ratio of the highest pollutant concentration within the partition to the safety limit, and the presence of high-risk equipment or personnel within the partition. A higher priority coefficient indicates that the partition requires prioritized and intensified cleanup.

[0029] The flexible execution scheduling module is responsible for transforming the abstract partition topology output by the dynamic partitioning decision module into specific execution unit scheduling instructions. Please refer to the appendix. Figure 3 With appendix Figure 4 This module includes an execution unit resource library, a task-resource matching unit, and a path planning and collaborative control unit. The execution unit resource library is a real-time updated database that registers and manages all available purification execution units within the system. These units are divided into two main categories: fixed units and mobile units. Fixed units include fresh air handling units installed in air conditioning systems, fixed high-efficiency air filter units embedded in ceilings or side walls, and ceiling-mounted circulating fans used to promote local air circulation. Mobile units mainly refer to mobile air purification robots equipped with autonomous mobile chassis, high-efficiency filter modules, and fans. Each execution unit has an attribute record in the resource library, including: a unique identifier, unit type, real-time location coordinates, rated purification airflow, effective radius of action, movement speed for mobile robots, current working status (idle, working, charging, faulty), and remaining energy percentage.

[0030] The task-resource matching unit receives the virtual purification partition topology and its complete attribute list from the dynamic partition decision module. For each virtual purification partition in the list, the unit first calculates the total purification airflow required to purify the partition to the target level based on the partition's volume, current average pollutant concentration, and the target cleanliness concentration to be achieved, combined with an empirical air exchange rate model. This is a dynamic requirement value. Subsequently, the task-resource matching unit initiates the resource matching process. Its core is to run a hybrid matching algorithm combining the Hungarian algorithm and a greedy strategy, aiming to allocate suitable combinations of purification execution units to all virtual purification partitions. The algorithm treats partitions as tasks and execution units as resources. First, the algorithm creates a candidate execution unit list for each partition, where the units in the list are those whose effective radius of action can cover most of the partition's area. Then, the algorithm constructs a cost matrix, where the rows of the matrix represent partitions and the columns represent candidate execution units. The value of each element in the matrix represents the cost of assigning that unit to perform the purification task for that partition. Cost calculation considers multiple factors: for fixed units, costs mainly include startup energy consumption and additional energy consumption during operation; for mobile robots, costs include the energy consumption of the estimated path length from their current location to the geometric center of the partition, startup energy consumption, and operating energy consumption. As a key optimization design, when assigning mobile air-purifying robots to virtual purification zones, the matching algorithm not only calculates the distance cost of the robot reaching the partition center but also includes the estimated pollutant concentration along the robot's required travel path as an additional cost item. Specifically, the system estimates the concentration at each point along the robot's planned path based on a dynamic pollutant concentration scalar field, assigning a discount factor (such as multiplying by a positive coefficient less than 1) to the path cost in high-concentration areas. This means that when the path passes through a high-concentration area, the overall cost is calculated after the discount, thus guiding the system to prioritize scheduling robots that can purify more pollutants along the way. This implies that scheduling a robot that needs to cross a polluted area will reduce its overall cost, thereby guiding the system to prioritize scheduling robots that can also purify the surrounding pollution on their way to the target partition, improving the overall purification benefit of a single scheduling. The Hungarian algorithm is used to find an allocation scheme that minimizes the total matching cost at the global level. The greedy strategy is used to handle real-time requirements, and performs rapid local reallocation when a new partition is urgently generated or the cell state changes abruptly.

[0031] The path planning and collaborative control unit specifically handles the scheduling details of the mobile air-purifying robot. This unit receives instructions from the task-resource matching unit, which include the target robot identifier and the geometric center coordinates or entrance coordinates of the target virtual purification zone. This unit first obtains the latest static map information from the environmental situational awareness module, including the outlines of walls and fixed equipment, as well as information on dynamic obstacles, the real-time positions and planned paths of other robots, and the real-time positions and predicted trajectories of personnel. Path planning employs an algorithm that integrates the dynamic window method and the time elastic band method. The dynamic window method samples multiple feasible velocity commands in the robot's velocity space, simulates the robot's motion trajectory over a short period, and evaluates whether each trajectory will collide, approach the target, and whether the velocity is smooth, thereby selecting the optimal instantaneous velocity command. The time elastic band method optimizes the entire initial path, allowing points on the path to move elastically in time and space to absorb the movement of dynamic obstacles, ultimately generating a collision-free, time-optimal, and smooth final path. After planning is completed, the path point sequence is sent to the corresponding mobile robot controller.

[0032] The collaborative control function targets multiple execution units that may be assigned within the same virtual purification zone. For example, a zone might simultaneously be assigned a ceiling-mounted circulating fan and a mobile air-purifying robot. The path planning and collaborative control unit needs to coordinate their operating parameters. Based on the zone's shape, the location of the pollution source, and the estimated airflow field, this unit sets recommended fan speed levels, air delivery angles for adjustable-direction units, and start / stop sequences for each execution unit within the zone. The core principle is to ensure that the airflow generated by different units coordinates to form a directional airflow pattern that facilitates pollutant discharge or dilution, avoiding airflow collisions and cancellations. For example, the mobile robot can be instructed to deliver airflow upwind of the pollution source, directing pollutants towards the fixed exhaust unit.

[0033] The global collaborative optimization module is the intelligent central hub for the system to achieve long-term optimal operation. Please refer to the appendix. Figure 5This module includes a multi-objective optimization model, an online parameter calibrator, and a system performance evaluation and feedback unit. The multi-objective optimization model constructs a scalar function characterizing the overall system operating cost. This function is a weighted sum of three cost components: purification efficiency cost, response time cost, and energy consumption cost. Purification efficiency cost is quantified as follows: the system sets a theoretical pollutant concentration decrease rate curve for each virtual purification zone based on its initial concentration and target concentration. Within each optimization cycle, the deviation between the actual concentration decrease rate and the theoretical rate for all zones is calculated, and the sum of the squares of these deviations is taken as the purification efficiency cost. Response time cost is quantified as follows: the time delay from the moment the pollution source is identified until the concentration in the affected risk area drops below the safety limit is recorded, and the sum of the time delays for all such events is taken as the response time cost. Energy consumption cost is quantified as follows: the instantaneous power of all fixed and mobile execution units in operation is collected in real time, the power is integrated over time, and the sum is obtained as the total energy consumption cost for that cycle. The final expression of the multi-objective optimization model is: Where J is the total system cost, Where M represents the purification efficiency cost, and M represents the total number of virtual partitions. and These represent the actual and target pollutant concentration decrease rates for zone m, respectively. The response time cost is E, and the total number of pollution events is E. Let e ​​be the delay time from the identification of the e-th event to the achievement of the key point; Where U represents the energy consumption cost, and U represents the total number of execution units in operation. Its instantaneous power. These are three positive weighting coefficients, and their relative magnitudes determine the system's preference for different objectives.

[0034] The online parameter calibrator acts as a bridge between the model and the front-end execution module. It is embedded in key decision-making stages of the dynamic partitioning decision module and the flexible execution scheduling module. The specific parameters calibrated include the weight coefficients in the edge weight calculation formula used by the partition topology solution unit in the dynamic partitioning decision module. The weighting coefficients for distance cost, in-transit purification benefit cost, and energy consumption cost in the cost calculation function of the task-resource matching algorithm in the flexible execution scheduling module. The online parameter calibrator operates on a minute-by-minute cycle. At the beginning of each cycle, it obtains historical performance data from the system performance evaluation and feedback unit for the past cycle, including the actual total cost J and the contribution of each component cost. The calibrator employs an online learning strategy based on gradient descent. It changes a parameter to be calibrated by a small perturbation, observes the trend of system cost J within the next very short time window, and estimates the impact of this parameter on the system cost J. The gradient of J is calculated. Then, the value of this parameter is adjusted in the direction that makes J decrease. This process is performed sequentially or simultaneously on all parameters to be calibrated, causing the system's decision-making behavior to continuously evolve in the direction of reducing the overall cost.

[0035] The system performance evaluation and feedback unit is responsible for continuously monitoring and evaluating the system's operational effectiveness. This unit obtains historical pollutant concentration variation curves for each virtual purification zone from the environmental situation awareness module, actual energy consumption data from the execution unit resource library and each unit controller, and the overall task completion time from the event log. Based on this data, the unit calculates a series of key performance indicators, such as the average purification efficiency within a cycle, defined as the ratio of total pollutant removal to time, the average energy consumption per unit of purification, and the average response delay for pollution events. These indicators are not only used to generate system operation reports but, more importantly, serve as performance feedback signals, packaged and sent to the online parameter calibrator, thus forming a complete "perception-decision-execution-evaluation-optimization" closed loop. Through this closed-loop mechanism, the global collaborative optimization module enables the system to continuously learn environmental characteristics and adaptively adjust decision parameters, thereby approaching the Pareto optimal frontier under multiple objectives in long-term operation.

[0036] The system has preset multiple typical scenario modes to adapt to different operational needs. For example, in the daily monitoring mode, the system weight coefficient is set to... medium, Lower The weighting is relatively high, emphasizing energy-saving operation while maintaining basic cleanliness. In emergency pollution response modes, such as simulated leak drills, the weighting coefficient switches to [a different value]. Very high, Very high, The level is very low, and the system suppresses and removes pollution as quickly as possible, regardless of energy consumption. In energy-saving operation modes at night or unattended, the weighting factor is set to... Lower Very low, The system operates at a very low level, maintaining only minimal monitoring and background purification. Users can manually select scene modes through the human-machine interface of the central control panel, or the system can automatically trigger mode switching based on preset schedules such as work schedules, night schedules, or by analyzing the frequency of historical events. This scene adaptability further enhances the system's practicality and intelligence.

[0037] The workflow of this invention's system is a highly coordinated cyclical process. After system power-on initialization, the environmental situation awareness module begins continuous operation, constructing and updating the dynamic pollution situation field. When no significant pollution is detected, the dynamic zoning decision module may treat the entire environment as a single zone or maintain a basic zoning structure, while the flexible execution scheduling module schedules a small number of units for background purification. Once the pollution source and risk zone identification unit detects a new active pollution source, the dynamic zoning decision module responds immediately, recalculating and generating a virtual purification zone topology containing the new pollution source within seconds. The new zoning instructions are then rapidly sent to the flexible execution module.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent zoned purification control system for experimental environments, characterized in that, include: The environmental situation awareness module is used to collect and fuse multi-dimensional state data in the experimental environment in real time to construct a dynamic pollution situation field. The dynamic zoning decision module is used to calculate and generate a virtual purification zone topology in real time based on the dynamic pollution situation field. The flexible execution scheduling module is used to assign tasks and plan action paths to the purification execution units according to the topology of the virtual purification partition. The global collaborative optimization module is used to establish and solve an optimization model with purification efficiency, response speed and system energy consumption as comprehensive objectives, and to perform online calibration and collaborative optimization of the operating parameters of the dynamic partition decision module and the flexible execution scheduling module. The environmental situation awareness module includes a distributed sensor network, a personnel and equipment activity tracking unit, and a data fusion and field construction unit. The distributed sensor network consists of multiple types of sensor nodes deployed in the experimental environment. The sensor types include at least laser particle counters, multi-gas component detectors, temperature and humidity sensors, and micro-differential pressure sensors. Each sensor node synchronously collects pollutant concentration, gas composition, temperature, humidity, and air pressure data at its location at a sampling frequency of no less than once per second. The personnel and equipment activity tracking unit tracks the location coordinates and movement trajectory of experimental personnel, mobile devices, and materials in real time through ultra-wideband positioning base stations and tags deployed in the environment. The data fusion and field construction unit receives raw data streams from the distributed sensor network and the personnel and equipment activity tracking unit. It uses a multi-source data spatiotemporal registration algorithm based on Kalman filtering to unify asynchronous and heterogeneous sensor data into the same spatiotemporal coordinate system. It then uses the Kriging space interpolation algorithm to calculate the pollutant concentration distribution at locations not directly monitored throughout the space in real time, based on the registered discrete point monitoring data and combined with the three-dimensional geometric model of the experimental environment and preset airflow organization parameters. This constructs a dynamic pollutant concentration scalar field covering the entire environment. The preset airflow organization parameters include the main airflow velocity vector field estimated based on micro-differential pressure sensor data and a simplified computational fluid dynamics model. The dynamic zoning decision module includes a pollution source and risk zone identification unit, a zoning topology calculation unit, and a zoning attribute definition unit. The pollution source and risk area identification unit interfaces with the dynamic pollution concentration scalar field and personnel activity trajectory data output by the data fusion and field construction unit. By setting concentration gradient threshold and time change rate threshold, it automatically identifies areas with rapidly rising concentrations and marks them as active pollution sources. At the same time, based on the real-time location of personnel and key equipment, combined with preset safety concentration limits, it delineates risk areas that need to be protected. The partitioned topology solving unit takes the identified pollution sources and risk zones as input, uses the physical space as nodes in graph theory, and uses the Euclidean distance, connectivity, and estimated airflow influence intensity between nodes as edge weights to construct an undirected weighted graph of the environmental space. It then runs a pollution diffusion blocking path search program based on an improved Dijkstra algorithm. The improvement lies in that, in the relaxation operation of the standard Dijkstra algorithm, not only is the cumulative path weight considered, but a dynamic penalty factor is also introduced. This dynamic penalty factor is proportional to the real-time pollutant concentration at the current path endpoint node, prioritizing the exploration of blocking paths in high-concentration areas. The program starts with each active pollution source and uses all risk zones as potential endpoints, searching and marking the critical paths and boundaries in the weighted graph that can most effectively block the diffusion of pollutants from the source to the risk zone. These marked paths and boundaries are spatially connected end-to-end, forming closed virtual boundaries, thereby dynamically dividing the environment into multiple mutually isolated virtual purification zones. The partition attribute definition unit assigns control attributes to each generated virtual purification partition. The attributes include at least the current average pollutant concentration, the highest pollutant concentration, the area and volume of the partition, the identification of the equipment and personnel contained in the partition, and the partition purification priority coefficient calculated based on the pollution source intensity and risk level.

2. The intelligent zoned purification control system for experimental environments according to claim 1, characterized in that, When the partitioned topology solving unit runs the pollution diffusion barrier path search program, the weight of the edges in the constructed undirected weighted graph is determined by three factors: the first factor is the physical distance between nodes; the second factor is the airflow intensity from upstream to downstream nodes estimated by the data fusion and field construction unit; and the third factor is whether there are physical barriers or wind curtains between nodes. The final weight of the edge is the weighted product of these three factors, calculated using the following formula: ,in, This represents the edge weight from node i to node j. Let be the Euclidean distance from node i to node j. Let ϵ be the estimated airflow intensity from node i to node j, where ϵ is a local constant. As a barrier factor, These are the weighting coefficients for distance, airflow, and obstruction, respectively. The global collaborative optimization module dynamically adjusts the parameters.

3. The intelligent zoned purification control system for experimental environments according to claim 1, characterized in that, The flexible execution scheduling module includes an execution unit resource library, a task-resource matching unit, and a path planning and collaborative control unit. The execution unit resource library registers and manages all available purification execution units in the system, including fixed fresh air units, fixed high-efficiency air filters, ceiling-mounted circulating fans, and mobile air purification robots with autonomous movement capabilities. The attribute records of each execution unit in the resource library include its unique identifier, type, real-time location coordinates, rated purification air volume, effective radius of action, movement speed, current working status, and remaining energy. The task-resource matching unit receives the virtual purification partition topology and its attributes from the dynamic partition decision module. For each virtual purification partition, it calculates the required total purification air volume based on the partition's spatial geometry, purification priority coefficient, and the target cleanliness level to be achieved. Based on the real-time status of the execution unit resource library, it uses a hybrid matching algorithm combining the Hungarian algorithm and a greedy strategy to allocate a suitable combination of purification execution units to each partition.

4. The intelligent zoned purification control system for experimental environments according to claim 3, characterized in that, When the task-resource matching unit assigns mobile air purification robots to virtual purification zones, the matching algorithm not only considers the distance between the robot's real-time position and the zone center, but also includes the estimated pollutant concentration on the path required for the robot to reach the zone as an additional cost item in the calculation. A discount coefficient is assigned to the cost of high-concentration areas on the path, and the discount coefficient is a positive coefficient less than 1.

5. The intelligent zoned purification control system for experimental environments according to claim 3, characterized in that, The path planning and collaborative control unit is specifically designed for mobile air purification robots. Based on the instructions issued by the task-resource matching unit, this unit plans the optimal travel path for mobile robots that need to go to a designated area to perform tasks. The path planning adopts an algorithm that integrates dynamic window method and time elastic band. This algorithm not only considers static obstacles, but also incorporates the activity trajectories of other mobile robots and personnel in real time as dynamic obstacles for obstacle avoidance calculation. At the same time, this unit generates collaborative control instructions to coordinate the wind speed, wind direction and start-stop sequence of multiple fixed and mobile execution units in the same area.

6. The intelligent zoned purification control system for experimental environments according to claim 1, characterized in that, The global collaborative optimization module includes a multi-objective optimization model, an online parameter calibrator, and a system performance evaluation and feedback unit. The multi-objective optimization model constructs a mathematical function with the goal of minimizing the overall system operating cost. This function is a weighted sum of purification efficiency cost, response time cost, and energy consumption cost. The online parameter calibrator is embedded in the key decision-making stages of the dynamic partitioning decision module and the flexible execution scheduling module. Specifically, it includes the edge weight calculation formula of the pollution diffusion blocking path search algorithm in the dynamic partitioning decision module and the preference weight of the task-resource matching algorithm in the flexible execution scheduling module. The online parameter calibrator receives historical performance data from the system performance evaluation and feedback unit at a time interval of minutes and dynamically adjusts the above-mentioned internal algorithm parameters using an online learning strategy based on gradient descent. The system performance evaluation and feedback unit continuously monitors the pollutant concentration change curves of each virtual purification zone, the actual energy consumption data of each execution unit, and the overall task completion time. It calculates the periodic purification efficiency, energy consumption per unit of purification volume, and average response delay index, and inputs these indexes as performance feedback to the online parameter calibrator.

7. The intelligent zoned purification control system for experimental environments according to claim 6, characterized in that, The purification efficiency cost is quantified by the sum of squares of the deviations between the actual pollutant concentration decrease rate and the target decrease rate in all virtual purification zones; the response time cost is quantified by the time delay from the identification of a pollution event to the attainment of the target concentration at key points within the zone. The energy consumption cost is quantified by summing the power integrals of all stationary and mobile execution units in operation.

8. The intelligent zoned purification control system for experimental environments according to claim 6, characterized in that, The weight coefficients of the multi-objective optimization model are not fixed. The system presets multiple typical scenario modes, including daily monitoring mode, emergency pollution response mode, and energy-saving operation mode. Under different scenario modes, the weight coefficients of purification efficiency, response speed, and energy consumption in the global cost function have different preset combinations.

9. The intelligent zoned purification control system for experimental environments according to claim 1, characterized in that, The data fusion and field construction unit also estimates the main airflow velocity vector field based on micro differential pressure sensor data and a simplified computational fluid dynamics model.

10. The intelligent zoned purification control system for experimental environments according to claim 1, characterized in that, The optimization objective of the hybrid matching algorithm is to minimize the overall start-up, shutdown, and movement energy consumption of the execution unit while meeting the air volume requirements of the partition purification system, and to prioritize scheduling the available unit closest to the geometric center of the partition.

Images