Fume hood dynamic air volume regulation and control method and system based on risk prediction
By integrating multi-source data and machine learning to predict risks, and combining feedforward and feedback control, the optimal exhaust volume is dynamically calculated, which solves the problem of lag in the control logic of fume hoods and improves safety and energy efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-03-24
AI Technical Summary
The control logic of existing fume hood control systems lags behind risk events, resulting in blind spots in safety assurance and wasted energy, making it difficult to cope with complex and ever-changing operating conditions.
By fusing multi-source data and using machine learning models to predict risks, and combining feedforward and feedback control strategies, the optimal exhaust volume is dynamically calculated to achieve proactive prevention and rapid response.
It improved the safety level of the laboratory environment, reduced the energy consumption of the ventilation system, and achieved rapid response and long-term stable air volume control.
Smart Images
Figure CN121715397A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of intelligent control, in particular to a fume hood dynamic air volume regulation method and system based on risk prediction. BACKGROUND
[0002] In the field of laboratory environment safety and energy conservation and environmental protection, as a key equipment to protect the operating personnel from harmful pollutants, the performance of the control system of the fume hood is crucial.
[0003] At present, the common improvement technology is mostly based on the feedback control principle. For example, the variable air volume fume hood adjusts the air volume in real time to maintain the constant cabinet face wind speed by monitoring the window opening degree; a more advanced method will introduce a pollutant concentration sensor to increase the air volume for remediation after detecting the leakage exceeding the standard. These methods still belong to the post-response or static following mode in essence, that is, the air volume is adjusted according to the state change or the detected danger. Although it is optimized compared with the constant air volume control, the control logic is always lagging behind the occurrence of the risk event.
[0004] However, the above-mentioned prior art has inherent limitations. The core problem is the lag of the control logic, whether it is to maintain a constant face wind speed or to increase the air volume after detecting the pollutant exceeding the standard, which belongs to the passive response mode of post-remediation. This mode cannot predict risks, resulting in that the system response is always slower than the occurrence of the risk event, thereby there is a time blind area in safety protection. In order to make up for this defect, the system often needs to set a higher basic air volume as a buffer, which greatly reduces the energy saving effect. In addition, the feedback mechanism that simply relies on a single or a small number of parameters is difficult to cope with complex and variable actual operation conditions, which easily leads to frequent action of the control system, affecting the stability of the air volume and the comfort of the environment. SUMMARY
[0005] In view of the deficiencies of the prior art, the purpose of the present application is to provide a fume hood dynamic air volume regulation method and system based on risk prediction, which aims to solve the above-mentioned problems recorded in the prior art.
[0006] The first aspect of the present application is to provide a fume hood dynamic air volume regulation method based on risk prediction, which comprises: Real-time collection of multi-source data of the fume hood in the running process by a plurality of pre-deployed sensors, wherein the multi-source data includes operating parameters, environmental parameters and operating behavior data of the fume hood; Fusion processing of the collected multi-source data to generate a standardized feature vector for risk assessment, and construction of historical time series data corresponding to the standardized feature vector; The standardized feature vector and the historical time series data are input into a pre-trained risk prediction model, which outputs the probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future. A multi-objective optimization function is constructed with multiple indicators as optimization targets. The risk probability value or risk level is used as input to solve the multi-objective optimization function and dynamically calculate and output the optimal exhaust volume set value corresponding to the current time. Based on the optimal exhaust volume setpoint, a composite control strategy combining feedforward and feedback is used to generate control commands for controlling the fume hood.
[0007] According to one aspect of the above technical solution, the steps of fusing the collected multi-source data to generate a standardized feature vector for risk assessment and constructing historical time-series data corresponding to the standardized feature vector include: The collected multi-source data is standardized and aligned, and the timestamps are unified to form standardized feature vectors at the same time. By combining multiple standardized feature vectors within the current time and a preset historical time period in chronological order, a feature time series window for characterizing dynamic change trends is constructed, thus obtaining historical sequence data.
[0008] According to one aspect of the above technical solution, the step of combining multiple standardized feature vectors within the current moment and a preset historical time period in chronological order to construct a feature time series window for characterizing dynamic change trends and obtaining historical sequence data includes: Set the length of the historical time series window and the data sampling interval; Based on the current time, the standardized feature vectors corresponding to multiple time points are extracted sequentially in chronological order; The standardized feature vectors obtained from sampling are combined into a three-dimensional data structure to obtain historical sequence data.
[0009] According to one aspect of the above technical solution, the risk prediction model is a recurrent neural network or a temporal convolutional network suitable for processing time-series data; The steps of inputting the standardized feature vector and the historical time-series data into a pre-trained risk prediction model, and outputting the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future, include: The standardized feature vector and the historical time series data are input into a pre-trained risk prediction model. By analyzing the historical time series data, dynamic change features related to the risk of pollutant escape are extracted. The dynamic change features include change trends, change rates, and periodic patterns. Based on the dynamic change characteristics and combined with the standardized feature vector at the current moment, a continuous value representing the probability of future risks is calculated and output, or a discrete risk level signal is output.
[0010] According to one aspect of the above technical solution, the step of calculating and outputting a continuous value representing the probability of future risks, or outputting a discrete risk level signal, based on the dynamic change characteristics and combined with the standardized feature vector at the current moment, includes: The extracted dynamic change features are fused with the standardized feature vector at the current moment, and a comprehensive risk index is calculated and output based on the preset feature weights. Map the comprehensive risk index to a predefined output range; If the output is a continuous value, it is mapped to a risk probability value between 0 and 1; if the output is a discrete signal, the risk index is mapped to the corresponding risk level by setting a risk threshold range.
[0011] According to one aspect of the above technical solution, the step of fusing the extracted dynamic change features with the standardized feature vector at the current moment, and calculating and outputting a comprehensive risk index based on preset feature weights, includes: The dynamically changing feature vector is concatenated with the standardized feature vector at the current moment to form a fused feature vector; Each feature in the fused feature vector is multiplied by a weight coefficient predetermined through model training, and all weighted feature values are summed to calculate and output the final comprehensive risk index.
[0012] According to one aspect of the above technical solution, the step of generating control commands for controlling the fume hood using a composite control strategy combining feedforward and feedback, based on the optimal exhaust volume setpoint, includes: Based on real-time identification of high-risk actions by operators or rapid opening of windows, an initial airflow control command for rapid compensation is generated according to preset feedforward control rules. The difference between the optimal exhaust volume setpoint and the actual exhaust volume of the fume hood is used as input, and the PID controller calculates the result to generate a feedback air volume control command for error elimination. The initial air volume control command corresponding to the feedforward control and the feedback air volume control command corresponding to the feedback control are combined to output a control command for controlling the exhaust actuator in the fume hood.
[0013] A second aspect of the present invention is to provide a risk-prediction-based dynamic airflow control system for fume hoods, applied to the method described in the above-mentioned technical solution, the system comprising: The data acquisition module is used to collect multi-source data in real time during the operation of the fume hood through a variety of pre-deployed sensors. The multi-source data includes the fume hood's operating parameters, environmental parameters, and operational behavior data. The data fusion module is used to fuse the collected multi-source data, generate a standardized feature vector for risk assessment, and construct historical time-series data corresponding to the standardized feature vector. The risk prediction module is used to input the standardized feature vector and the historical time series data into a pre-trained risk prediction model and output the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future. The dynamic solution module is used to construct a multi-objective optimization function with multiple indicators as optimization objectives, and solve the multi-objective optimization function with the risk probability value or risk level as input, and dynamically calculate and output the optimal exhaust volume set value corresponding to the current time. The instruction generation module is used to generate control instructions for controlling the fume hood based on the optimal exhaust volume setting value and using a composite control strategy that combines feedforward and feedback.
[0014] A third aspect of the present invention is to provide a readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in the above-described technical solution.
[0015] A fourth aspect of the present invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the above technical solutions.
[0016] Compared with existing technologies, the advantages of the risk prediction-based dynamic airflow control method and system for fume hoods shown in this invention are as follows: This embodiment integrates multi-source sensing data and utilizes machine learning for proactive risk prediction, enabling the assessment of the likelihood of pollutant escape risks before they actually occur. This elevates control logic from reactive remediation to proactive prevention. Furthermore, by establishing an optimization function with multiple objectives—safety, energy efficiency, and comfort—for real-time decision-making, it ensures that, with personnel safety as the absolute priority, the optimal airflow rate matching the actual risk level at any given moment can be dynamically calculated. This completely avoids the enormous energy waste caused by maintaining high airflow rates year-round for safety reasons, as is common in traditional methods. In addition, the application of a feedforward-feedback composite control strategy combines rapid compensation for predictable disturbances with precise correction of system deviations, resulting in airflow regulation that possesses both rapid response capabilities and long-term stability. In summary, this embodiment not only significantly improves the environmental safety level of the laboratory but also substantially reduces the operating energy consumption of the ventilation system. Attached Figure Description
[0017] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 A flowchart illustrating the dynamic airflow control method for fume hoods based on risk prediction provided in an embodiment of the present invention; Figure 2 The structural block diagram of the dynamic air volume control system for fume hoods based on risk prediction provided in the embodiments of the present invention. Detailed Implementation
[0018] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be more thorough and complete.
[0019] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.
[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.
[0021] Example 1 The first embodiment of the present invention provides a method for dynamic airflow control of fume hoods based on risk prediction, the method comprising steps S10-S50: Step S10: Collect multi-source data in real time during the operation of the fume hood using a variety of pre-deployed sensors. The multi-source data includes the fume hood's operating parameters, environmental parameters, and operational behavior data.
[0022] Specifically, the operating parameters of the fume hood are acquired through sensors installed on the fume hood. For example, high-precision angle sensors monitor the real-time opening of the window sashes, and valve positioners monitor the valve opening of the exhaust valve to assess the duct resistance. Environmental parameters of the fume hood are monitored in real-time by gas sensors, such as VOCs sensors and dust sensors, installed inside the fume hood or exhaust duct, to detect the concentration of target pollutants. Operational behavior data is obtained by using infrared sensors, millimeter-wave radar, or vision sensors to perceive the operator's state and intentions, such as determining whether a person is working in front of the fume hood and identifying predefined high-risk actions such as "pouring," "heating," and "moving large containers."
[0023] Step S20: The collected multi-source data is fused to generate a standardized feature vector for risk assessment, and historical time series data corresponding to the standardized feature vector is constructed.
[0024] In this embodiment, the steps of fusing the collected multi-source data to generate a standardized feature vector for risk assessment and constructing historical time-series data corresponding to the standardized feature vector include: The collected multi-source data is standardized and aligned, and the timestamps are unified to form standardized feature vectors at the same time. By combining multiple standardized feature vectors within the current time and a preset historical time period in chronological order, a feature time series window for characterizing dynamic change trends is constructed, thus obtaining historical sequence data.
[0025] The step of combining multiple standardized feature vectors within the current time and a preset historical time period in chronological order to construct a feature time series window for characterizing dynamic trends and obtaining historical sequence data includes: Set the length of the historical time series window and the data sampling interval; Based on the current time, the standardized feature vectors corresponding to multiple time points are extracted sequentially in chronological order; The standardized feature vectors obtained from sampling are combined into a three-dimensional data structure to obtain historical sequence data.
[0026] Specifically, in this embodiment, the process of constructing a feature time series window is the core of realizing the transition from static perception to dynamic prediction. The data collected by the original sensors is instantaneous and isolated, which can only reflect the state at a certain point in time and cannot capture the evolution trend of risk. By setting the length of the historical time series window and the data sampling interval, an observation time scale for analyzing risk dynamics is defined.
[0027] More specifically, the above process relies on the sliding window algorithm, which takes the current time as a reference and extracts the standardized feature vector corresponding to each time within a specific time period in the past according to the set sampling interval. Finally, these multiple standardized feature vectors arranged in chronological order are combined into a three-dimensional data structure to obtain historical sequence data.
[0028] Step S30: Input the standardized feature vector and the historical time series data into the pre-trained risk prediction model, and output the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future.
[0029] In this embodiment, the risk prediction model is a recurrent neural network or a temporal convolutional network suitable for processing time-series data; The steps of inputting the standardized feature vector and the historical time-series data into a pre-trained risk prediction model, and outputting the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future, include: The standardized feature vector and the historical time series data are input into a pre-trained risk prediction model. By analyzing the historical time series data, dynamic change features related to the risk of pollutant escape are extracted. The dynamic change features include change trends, change rates, and periodic patterns. Based on the dynamic change characteristics and combined with the standardized feature vector at the current moment, a continuous value representing the probability of future risks is calculated and output, or a discrete risk level signal is output.
[0030] The step of calculating and outputting a continuous value representing the probability of future risk, or outputting a discrete risk level signal, based on the dynamic change characteristics and the standardized feature vector at the current moment, includes: The extracted dynamic change features are fused with the standardized feature vector at the current moment, and a comprehensive risk index is calculated and output based on the preset feature weights. Map the comprehensive risk index to a predefined output range; If the output is a continuous value, it is mapped to a risk probability value between 0 and 1; if the output is a discrete signal, the risk index is mapped to the corresponding risk level by setting a risk threshold range.
[0031] In addition, the step of fusing the extracted dynamic change features with the standardized feature vector at the current moment, and calculating and outputting a comprehensive risk index based on preset feature weights, includes: The dynamically changing feature vector is concatenated with the standardized feature vector at the current moment to form a fused feature vector; Each feature in the fused feature vector is multiplied by a weight coefficient predetermined through model training, and all weighted feature values are summed to calculate and output the final comprehensive risk index.
[0032] Specifically, in this embodiment, the pre-trained risk prediction model typically employs a machine learning model suitable for processing time-series data, such as a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN). The structured time-series data samples containing historical and current information, constructed as described above, are imported into the model. Internally, the model uses its multi-layer neural network structure to first perform in-depth analysis of the historical time-series data, automatically extracting dynamic feature patterns strongly correlated with risk. These patterns include, but are not limited to: whether the trend of pollutant concentration change is stable, linearly increasing, or accelerating; the rate of change of window opening; and short-term periodic patterns that certain operational actions may exhibit.
[0033] Subsequently, the dynamic patterns learned from historical data are comprehensively analyzed and weighted for decision-making based on the latest standardized feature vectors at the current moment. Specifically, based on the complex nonlinear relationships learned during training, it determines what future outcome the current state will lead to under what historical trends. Ultimately, the model outputs a quantitative prediction value: for example, a continuous probability value between 0 and 1, representing the probability of pollutant emissions occurring within a specific time period in the future; or a discrete risk level signal, providing accurate input for subsequent control decisions.
[0034] Step S40: Construct a multi-objective optimization function with multiple indicators as optimization targets, use the risk probability value or risk level as input, solve the multi-objective optimization function, and dynamically calculate and output the optimal exhaust volume setting value corresponding to the current time.
[0035] Specifically, the first step is to construct an optimization model containing multiple sub-objective functions. The safety objective function acts as a constraint, designed to ensure that the exhaust volume is sufficient to suppress the predicted risk probability below a preset safety threshold; it typically exhibits a monotonically increasing risk level function. The energy-saving objective function is directly related to the exhaust volume, aiming to minimize it and thus reduce fan energy consumption. Additionally, a comfort / stability objective function is introduced, penalizing drastic changes in exhaust volume within adjacent control cycles to maintain indoor airflow stability and acoustic comfort. These objective functions are integrated into a complete, solvable multi-objective optimization function using weighted summation or constraint methods.
[0036] During the solution process, real-time risk prediction values are used as the primary input to dynamically calculate the optimal solution for the current moment. For example, when the predicted risk is determined to be low, the optimization algorithm, while meeting basic safety constraints, assigns higher weight to the energy-saving target and outputs a lower airflow value close to the minimum safety requirement. Once the predicted risk increases, the algorithm adaptively adjusts the weights, and the safety target instantly takes precedence, driving a significant increase in output airflow to prioritize safety. Through this continuous dynamic optimization, the airflow control of the fume hood has achieved a transformation from passive response to active prediction, and from energy waste to on-demand supply, maximizing energy efficiency while ensuring safety.
[0037] Step S50: Based on the optimal exhaust volume setting value, a composite control strategy combining feedforward and feedback is used to generate control commands for controlling the fume hood.
[0038] In this embodiment, the step of generating control commands for controlling the fume hood using a composite control strategy combining feedforward and feedback, based on the optimal exhaust volume setpoint, includes: Based on real-time identification of high-risk actions by operators or rapid opening of windows, an initial airflow control command for rapid compensation is generated according to preset feedforward control rules. The difference between the optimal exhaust volume setpoint and the actual exhaust volume of the fume hood is used as input, and the PID controller calculates the result to generate a feedback air volume control command for error elimination. The initial air volume control command corresponding to the feedforward control and the feedback air volume control command corresponding to the feedback control are combined to output a control command for controlling the exhaust actuator in the fume hood.
[0039] Specifically, feedforward control plays a predictive compensation role. Instead of waiting for errors to occur before acting, it outputs a control variable in advance based on real-time sensing of precursory disturbance signals that could cause sudden changes in airflow demand. These disturbance signals mainly fall into two categories: first, high-risk actions of operators detected by a behavior recognition system, such as recognizing the initial action of picking up a reagent bottle to pour it; and second, the rapid opening of windows detected by sensors. Once these signals are detected, the feedforward controller immediately generates a basic control command according to preset rules, causing the exhaust fan to accelerate in advance, thereby establishing an effective capture airflow in the early stages of pollutant diffusion.
[0040] Feedback control plays a crucial role in precise correction, forming a closed-loop control system. It continuously monitors the actual exhaust volume of the fume hood and compares it with the optimal exhaust volume setpoint, generating an error signal. This error signal is then input into the PID controller, which calculates a correction based on the current error, accumulated historical errors, and error trends. This correction effectively eliminates steady-state errors caused by factors such as duct pressure fluctuations and actuator precision, ensuring that the actual airflow consistently and accurately tracks the setpoint over a long period.
[0041] Ultimately, the control system superimposes the rapid compensation command from the feedforward control output and the precise correction command from the feedback control output to generate the final control command. Based on this control command, the variable frequency fan and damper in the fume hood are driven to move, which enables both an advanced response to sudden risks and long-term stability of the control effect, thus achieving a balance between response speed and control accuracy.
[0042] Compared with existing technologies, the risk prediction-based dynamic airflow control method for fume hoods shown in this embodiment has the following advantages: This embodiment integrates multi-source sensing data and utilizes machine learning for proactive risk prediction, enabling the assessment of the likelihood of pollutant escape risks before they actually occur. This elevates control logic from reactive remediation to proactive prevention. Furthermore, by establishing an optimization function with multiple objectives—safety, energy efficiency, and comfort—for real-time decision-making, it ensures that, with personnel safety as the absolute priority, the optimal airflow rate matching the actual risk level at any given moment can be dynamically calculated. This completely avoids the enormous energy waste caused by maintaining high airflow rates year-round for safety reasons, as is common in traditional methods. In addition, the application of a feedforward-feedback composite control strategy combines rapid compensation for predictable disturbances with precise correction of system deviations, resulting in airflow regulation that possesses both rapid response capabilities and long-term stability. In summary, this embodiment not only significantly improves the environmental safety level of the laboratory but also substantially reduces the operating energy consumption of the ventilation system.
[0043] Example 2 Please see Figure 2 The second embodiment of the present invention provides a risk prediction-based dynamic airflow control system for fume hoods, applied to the method described in any of the above embodiments, the system comprising: The data acquisition module 10 is used to collect multi-source data of the fume hood in real time during operation through a variety of pre-deployed sensors. The multi-source data includes the fume hood's operating parameters, environmental parameters, and operational behavior data. The data fusion module 20 is used to fuse the collected multi-source data, generate a standardized feature vector for risk assessment, and construct historical time series data corresponding to the standardized feature vector. Risk prediction module 30 is used to input the standardized feature vector and the historical time series data into a pre-trained risk prediction model and output the risk probability value or risk level of pollutant emission from the fume hood within a predetermined time window in the future. The dynamic solution module 40 is used to construct a multi-objective optimization function with multiple indicators as optimization objectives, and solve the multi-objective optimization function with the risk probability value or risk level as input, and dynamically calculate and output the optimal exhaust volume set value corresponding to the current time. The instruction generation module 50 is used to generate control instructions for controlling the fume hood based on the optimal exhaust volume setting value and using a composite control strategy that combines feedforward and feedback.
[0044] Compared with existing technologies, the advantages of using the risk prediction-based dynamic airflow control system for fume hoods shown in this embodiment are as follows: This embodiment integrates multi-source sensing data and utilizes machine learning for proactive risk prediction, enabling the assessment of the likelihood of pollutant escape risks before they actually occur. This elevates control logic from reactive remediation to proactive prevention. Furthermore, by establishing an optimization function with multiple objectives—safety, energy efficiency, and comfort—for real-time decision-making, it ensures that, with personnel safety as the absolute priority, the optimal airflow rate matching the actual risk level at any given moment can be dynamically calculated. This completely avoids the enormous energy waste caused by maintaining high airflow rates year-round for safety reasons, as is common in traditional methods. In addition, the application of a feedforward-feedback composite control strategy combines rapid compensation for predictable disturbances with precise correction of system deviations, resulting in airflow regulation that possesses both rapid response capabilities and long-term stability. In summary, this embodiment not only significantly improves the environmental safety level of the laboratory but also substantially reduces the operating energy consumption of the ventilation system.
[0045] Example 3 A third embodiment of the present invention provides a readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in any of the above embodiments.
[0046] Example 4 A fourth embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the methods described in the above embodiments.
[0047] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0048] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.
Claims
1. A method for dynamic airflow control of fume hoods based on risk prediction, characterized in that, The method includes: Multiple data sources are collected in real time by pre-deployed sensors during the operation of the fume hood. These data sources include the fume hood's operating parameters, environmental parameters, and operational behavior data. The collected multi-source data are fused to generate a standardized feature vector for risk assessment, and historical time-series data corresponding to the standardized feature vector is constructed. The standardized feature vector and the historical time series data are input into a pre-trained risk prediction model, which outputs the probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future. A multi-objective optimization function is constructed with multiple indicators as optimization targets. The risk probability value or risk level is used as input to solve the multi-objective optimization function and dynamically calculate and output the optimal exhaust volume set value corresponding to the current time. Based on the optimal exhaust volume setpoint, a composite control strategy combining feedforward and feedback is used to generate control commands for controlling the fume hood.
2. The method for dynamic airflow control of fume hoods based on risk prediction according to claim 1, characterized in that, The steps of fusing the collected multi-source data to generate standardized feature vectors for risk assessment, and constructing historical time-series data corresponding to the standardized feature vectors, include: The collected multi-source data is standardized and aligned, and the timestamps are unified to form standardized feature vectors at the same time. By combining multiple standardized feature vectors within the current time and a preset historical time period in chronological order, a feature time series window for characterizing dynamic change trends is constructed, thus obtaining historical sequence data.
3. The method for dynamic airflow control of fume hoods based on risk prediction according to claim 2, characterized in that, The steps of combining multiple standardized feature vectors within the current timeframe and a preset historical timeframe in chronological order to construct a feature time series window for characterizing dynamic trends and obtaining historical sequence data include: Set the length of the historical time series window and the data sampling interval; Based on the current time, the standardized feature vectors corresponding to multiple time points are extracted sequentially in chronological order; The standardized feature vectors obtained from sampling are combined into a three-dimensional data structure to obtain historical sequence data.
4. The method for dynamic airflow control of fume hoods based on risk prediction according to claim 1, characterized in that, The risk prediction model is a recurrent neural network or a temporal convolutional network suitable for processing time-series data. The steps of inputting the standardized feature vector and the historical time-series data into a pre-trained risk prediction model, and outputting the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future, include: The standardized feature vector and the historical time series data are input into a pre-trained risk prediction model. By analyzing the historical time series data, dynamic change features related to the risk of pollutant escape are extracted. The dynamic change features include change trends, change rates, and periodic patterns. Based on the dynamic change characteristics and combined with the standardized feature vector at the current moment, a continuous value representing the probability of future risks is calculated and output, or a discrete risk level signal is output.
5. The dynamic airflow control method for fume hoods based on risk prediction according to claim 4, characterized in that, Based on the dynamic change characteristics and combined with the standardized feature vector at the current moment, the steps of calculating and outputting a continuous value representing the probability of future risk, or outputting a discrete risk level signal, include: The extracted dynamic change features are fused with the standardized feature vector at the current moment, and a comprehensive risk index is calculated and output based on the preset feature weights. Map the comprehensive risk index to a predefined output range; If the output is a continuous value, it is mapped to a risk probability value between 0 and 1; if the output is a discrete signal, the risk index is mapped to the corresponding risk level by setting a risk threshold range.
6. The dynamic airflow control method for fume hoods based on risk prediction according to claim 5, characterized in that, The steps of fusing the extracted dynamic change features with the standardized feature vector at the current moment, and calculating and outputting a comprehensive risk index based on preset feature weights, include: The dynamically changing feature vector is concatenated with the standardized feature vector at the current moment to form a fused feature vector; Each feature in the fused feature vector is multiplied by a weight coefficient predetermined through model training, and all weighted feature values are summed to calculate and output the final comprehensive risk index.
7. The method for dynamic airflow control of fume hoods based on risk prediction according to any one of claims 1-6, characterized in that, The step of generating control commands for controlling the fume hood using a composite control strategy combining feedforward and feedback, based on the optimal exhaust volume setpoint, includes: Based on real-time identification of high-risk actions by operators or rapid opening of windows, an initial airflow control command for rapid compensation is generated according to preset feedforward control rules. The difference between the optimal exhaust volume setpoint and the actual exhaust volume of the fume hood is used as input, and the PID controller calculates the result to generate a feedback air volume control command for error elimination. The initial air volume control command corresponding to the feedforward control and the feedback air volume control command corresponding to the feedback control are combined to output a control command for controlling the exhaust actuator in the fume hood.
8. A dynamic airflow control system for fume hoods based on risk prediction, characterized in that, The system, applicable to the method of any one of claims 1-7, comprises: The data acquisition module is used to collect multi-source data in real time during the operation of the fume hood through a variety of pre-deployed sensors. The multi-source data includes the fume hood's operating parameters, environmental parameters, and operational behavior data. The data fusion module is used to fuse the collected multi-source data, generate a standardized feature vector for risk assessment, and construct historical time-series data corresponding to the standardized feature vector. The risk prediction module is used to input the standardized feature vector and the historical time series data into a pre-trained risk prediction model and output the risk probability value or risk level of pollutant escape from the fume hood within a predetermined time window in the future. The dynamic solution module is used to construct a multi-objective optimization function with multiple indicators as optimization objectives, and solve the multi-objective optimization function with the risk probability value or risk level as input, and dynamically calculate and output the optimal exhaust volume set value corresponding to the current time. The instruction generation module is used to generate control instructions for controlling the fume hood based on the optimal exhaust volume setting value and using a composite control strategy that combines feedforward and feedback.
9. A readable storage medium having computer instructions stored thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-7.