Clean room dynamic energy consumption management system and method based on AI

Abstract

The invention relates to the technical field of energy consumption management, and particularly discloses an AI-based clean room dynamic energy consumption management system and method. According to the method, a deep learning algorithm is introduced to carry out deep interaction analysis on clean room indoor air quality parameter time sequence data and environment change driving event data such as the number of indoor personnel, the external environment temperature, the room door opening and closing state and the equipment power, so that intelligent reasoning of the dynamic evolution process of the clean room indoor air quality is realized; according to the method, the air quality parameter prediction result is compared with the preset air quality standard range so as to predict the air quality change in the future short time under the current driving event, and then the short-time air quality parameter prediction result is compared with the preset air quality standard range so as to carry out feedforward adjustment on the HVAC system operation parameters of the clean room according to the difference between the short-time air quality parameter prediction result and the preset air quality standard range. According to the method, environmental changes can be actively predicted, smooth pre-adjustment is conducted in advance, the air quality in the clean room is kept within the preset standard range, meanwhile, unnecessary energy waste is remarkably reduced, and the environmental stability is improved.

Description

Technical Field

[0001] The present application relates to the field of energy consumption management technology, and more specifically, to an AI-based clean room dynamic energy consumption management system and method. Background Art

[0002] Cleanrooms, also known as dust-free or clean rooms, are essential infrastructure for high-tech industries and cutting-edge scientific research, including semiconductor manufacturing, biopharmaceuticals, aerospace, and precision instrument production. Their core function is to control airborne particulate matter, temperature, humidity, pressure, microbial contaminants, and other contaminants within a specific space within strict, pre-set standards through specially designed architecture, equipment, and operational management. This ensures product yield, experimental accuracy, and production process reliability. To achieve and maintain this highly clean environment, the cleanroom's heating, ventilation, and air conditioning (HVAC) system must operate continuously, diluting and removing contaminants generated within it through high-volume, high-frequency air circulation and filtration. This process consumes significant energy, making cleanroom operating costs, especially energy costs, a significant portion of total operating expenses. Therefore, developing efficient, dynamic energy management solutions, while ensuring consistent cleanliness standards, has become a crucial issue within the industry for reducing costs and achieving green and sustainable development.

[0003] In existing technology, cleanroom HVAC system control generally employs a feedback-based regulation strategy. This approach typically deploys sensors throughout the cleanroom to monitor key air quality parameters such as temperature, humidity, and particulate matter concentration in real time. When a sensor-detected parameter deviates from a preset threshold, a control system (such as a classic PID controller) responds by adjusting actuators such as fan speed and cooling coil valve opening, increasing or decreasing the HVAC system's operating intensity until the exceeded parameter is brought back within the specified range. However, this passive feedback control approach exhibits significant hysteresis, requiring the control system to react only after air quality deterioration has already occurred and been detected by the sensors. This can cause the cleanroom environment to experience brief contamination events, potentially impacting sensitive production or laboratory activities. Furthermore, to quickly correct deviations, large and drastic adjustments are often required. This operating mode not only easily leads to parameter overshoot and oscillation, disrupting environmental stability, but also causes transient spikes in energy consumption, resulting in unnecessary energy waste.

[0004] Therefore, we look forward to an optimized AI-based clean room dynamic energy consumption management method and system. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an AI-based clean room dynamic energy consumption management system and method, which introduces a deep learning algorithm to conduct in-depth interactive analysis of the time series data of clean room air quality parameters and environmental change driving event data such as the number of indoor people, external ambient temperature, door switch status, equipment power, etc., to achieve intelligent reasoning of the dynamic evolution process of clean room air quality, thereby predicting the air quality changes in the short term in the future under the current driving event, and then, by comparing the short-term air quality parameter prediction results with the preset air quality standard range, the HVAC system operating parameters of the clean room are feedforward adjusted according to the difference between the two. This method can actively foresee environmental changes and perform smooth pre-adjustment in advance to maintain the clean room air quality within the preset standard range, while significantly reducing unnecessary energy waste and improving environmental stability.

[0006] According to one aspect of the present application, a clean room dynamic energy consumption management method based on AI is provided, which includes: Obtain a time series of air quality parameters of the clean room within a preset time window including the current time point, wherein the air quality parameters include temperature, humidity, PM2.5 concentration, and carbon dioxide concentration; Acquire environmental change-driven event data of the clean room in real time, wherein the environmental change-driven event data includes the current number of people in the clean room, the door switch status, the external ambient temperature, and the operating power of the indoor equipment; Inputting the time series of air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain a short-term air quality parameter prediction result; An air quality adjustment instruction is generated based on a comparison between the short-term air quality parameter prediction result and a preset air quality standard range.

[0007] According to another aspect of the present application, an AI-based clean room dynamic energy consumption management system is provided, which includes: An air quality parameter acquisition module is used to obtain a time series of air quality parameters of the clean room within a preset time window including the current time point, wherein the air quality parameters include temperature, humidity, PM2.5 concentration and carbon dioxide concentration; A drive event data acquisition module is used to acquire the environmental change drive event data of the clean room in real time, wherein the environmental change drive event data includes the current number of people in the clean room, the door switch status, the external environment temperature, and the operating power of the indoor equipment; An air quality prediction module is configured to input the time series of air quality parameters within a preset time window including the current time point and the environmental change driving event data into an air quality AI prediction model to obtain a short-term air quality parameter prediction result; The air quality adjustment instruction generating module is used to generate an air quality adjustment instruction based on the comparison between the short-term air quality parameter prediction result and the preset air quality standard range.

[0008] Compared with the existing technology, the AI-based clean room dynamic energy consumption management system and method provided by this application introduces a deep learning algorithm to conduct in-depth interactive analysis of the time series data of clean room air quality parameters and environmental change driving event data such as the number of indoor people, external ambient temperature, door switch status, and equipment power, so as to achieve intelligent reasoning of the dynamic evolution process of clean room air quality, thereby predicting the air quality changes in the short term in the future under the current driving event, and then, by comparing the short-term air quality parameter prediction results with the preset air quality standard range, the HVAC system operating parameters of the clean room are feedforward adjusted according to the difference between the two. This method can proactively foresee environmental changes and perform smooth pre-adjustment in advance to maintain the clean room air quality within the preset standard range, while significantly reducing unnecessary energy waste and improving environmental stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of an AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0011] Figure 2 This is a data flow diagram of the AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0012] Figure 3 This is a flowchart of sub-step S3 of the AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0013] Figure 4 This is a flowchart of sub-step S33 of the AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0014] Figure 5This is a flowchart of sub-step S331 of the AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0015] Figure 6 This is a flowchart of sub-step S4 of the AI-based clean room dynamic energy consumption management method according to an embodiment of the present application.

[0016] Figure 7 This is a block diagram of an AI-based clean room dynamic energy management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0018] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0019] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0020] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0021] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0022] In response to the technical problems described in the above background technology, this application proposes an AI-based clean room dynamic energy consumption management method, which introduces a deep learning algorithm to conduct in-depth interactive analysis of the time series data of clean room air quality parameters and environmental change driving event data such as the number of indoor people, external ambient temperature, door switch status, and equipment power, so as to achieve intelligent reasoning of the dynamic evolution process of clean room air quality, thereby predicting the air quality changes in the short term in the future under the current driving event, and then, by comparing the short-term air quality parameter prediction results with the preset air quality standard range, the HVAC system operating parameters of the clean room are feedforward adjusted according to the difference between the two. This method can proactively foresee environmental changes and perform smooth pre-adjustment in advance to maintain the clean room air quality within the preset standard range, while significantly reducing unnecessary energy waste and improving environmental stability.

[0023] Figure 1 This is a flowchart of an AI-based clean room dynamic energy consumption management method according to an embodiment of the present application. Figure 2 Figure 1 is a data flow diagram of the clean room dynamic energy consumption management method based on AI according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the AI-based dynamic energy consumption management method for clean rooms includes the following steps: S1, obtaining the time series of air quality parameters of the clean room within a preset time window including the current time point, the air quality parameters including temperature, humidity, PM2.5 concentration and carbon dioxide concentration; S2, obtaining the environmental change driving event data of the clean room in real time, the environmental change driving event data including the current number of people in the clean room, the door switch status, the external ambient temperature, and the operating power of the indoor equipment; S3, inputting the time series of air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain a short-term air quality parameter prediction result; S4, generating an air quality adjustment instruction based on the comparison between the short-term air quality parameter prediction result and the preset air quality standard range.

[0024] In the aforementioned AI-based cleanroom dynamic energy management method, step S1 obtains a time series of cleanroom air quality parameters within a preset time window that includes the current time point. These air quality parameters include temperature, humidity, PM2.5 concentration, and carbon dioxide concentration. It should be understood that the cleanroom environment is continuously influenced by historical conditions (such as temperature inertia and the cumulative effect of pollutants). Instantaneous air quality parameters (such as temperature, humidity, PM2.5 concentration, and carbon dioxide concentration) at a single time point cannot reflect the dynamic evolution of the environment. Therefore, to construct a complete portrait of the dynamic evolution of the cleanroom environment, this application obtains a time series of cleanroom air quality parameters within a preset time window that includes the current time point, forming a data foundation that can reflect the recent historical dynamics of the indoor environment and achieve a basic understanding of air quality change trends. Specifically, first, a high-precision sensor network strategically deployed throughout the cleanroom continuously collects real-time data on temperature, humidity, PM2.5 concentration, and carbon dioxide concentration. The instantaneous value of each parameter is recorded at a fixed sampling frequency (e.g., 10 seconds) and associated with a timestamp to form a raw data stream. Then, starting from the current time point, the system backtracks for a preset length of time (e.g., 15 minutes), capturing all data points within that timeframe to construct a multidimensional parameter time series consisting of multiple time steps. This approach not only captures a static snapshot of the cleanroom environment at the current moment but also constructs a mathematical representation of the recent fluctuation frequency and amplitude of air quality parameters, providing fundamental input for studying the inherent dynamic characteristics and inertial patterns of the cleanroom environmental system.

[0025] During implementation, a monitoring network consisting of multiple high-precision sensors must be deployed throughout the cleanroom, including but not limited to temperature and humidity sensors, particulate matter detectors (for measuring PM2.5), and carbon dioxide concentration sensors. These devices must exhibit excellent response speed and measurement accuracy to ensure the reliability and representativeness of the collected data. Sensor installation locations should be strategically arranged based on factors such as the cleanroom's functional zoning, airflow organization, and the frequency of human activity to avoid blind spots or data distortion. For example, sampling points should be set near air supply and return vents, on work surfaces, and in areas with high human activity to comprehensively capture differences in air quality across different spatial locations.

[0026] Secondly, in terms of acquisition frequency, the system is set to measure and record each parameter at a fixed period (e.g., once every 10 seconds), forming a continuous data stream. This high-frequency sampling strategy not only helps capture subtle fluctuations in the environment, but also improves the sensitivity of subsequent analysis models to short-term disturbances. The data acquired at each sampling moment is accompanied by a precise timestamp to maintain timeline consistency when constructing time series later. In addition, all raw data is transmitted to the central data processing unit via wired or wireless communication protocols, which is responsible for storage, preliminary cleaning, and formatting, providing a unified data interface for subsequent time window capture.

[0027] When constructing a dynamic environmental profile at a specific point in time, the system uses that point as a baseline, looks back for a preset period of time (e.g., 15 minutes), and integrates all data points collected during this period to form a multidimensional parameter time series consisting of multiple time steps. The selection of this time window requires comprehensive consideration of factors such as the cleanroom's environmental response characteristics, the duration of interference sources, and the prediction or feedback delay required by the control system. For example, if the cleanroom's temperature control system responds slowly, the lookback time can be appropriately extended to cover a sufficiently long historical period. For rapidly changing pollutant concentrations, the window length can be shortened to focus on recent behavioral characteristics.

[0028] During this process, constructing a time series involves more than simply stacking historical data together. Rather, it ensures that parameter values within each time step originate from the same sampling period and exhibit temporal continuity and consistency. To achieve this, the system requires a robust time synchronization mechanism to ensure that all sensor nodes operate on the same time basis and automatically identify and correct time misalignments caused by network delays, device failures, and other factors. Furthermore, the raw data must undergo necessary quality control to eliminate outliers or missing values, and interpolation algorithms must be used to fill gaps when necessary to ensure the integrity and accuracy of the time series.

[0029] In the aforementioned AI-based cleanroom dynamic energy management method, step S2 acquires real-time environmental change-driven event data for the cleanroom. This data includes the current number of cleanroom occupants, door status, external ambient temperature, and indoor equipment operating power. It should be understood that since changes in cleanroom air quality are primarily triggered by discrete events (such as increased CO2 emissions from an influx of people, the introduction of external pollutants and heat and moisture exchange from door opening, and temperature and humidity fluctuations caused by changes in the operating status of equipment such as ventilation and air conditioning), relying solely on historical time-series data cannot respond to the immediate impact of external disturbances on the indoor environment. Therefore, this application integrates diverse environmental monitoring and sensing technologies to simultaneously collect the current number of cleanroom occupants, door status, external temperature, and equipment operating power in real time to comprehensively capture various environmental change-driven events that affect cleanroom air quality. Specifically, the current number of cleanroom occupants and door status are provided by the access control system, accurately recording the number of people entering and exiting the cleanroom and door opening and closing movements. The external ambient temperature is measured in real time by a temperature sensor located outside the cleanroom, reflecting changes in the outdoor thermal environment. Equipment operating power is captured through smart meters connected to each device, accurately monitoring the energy consumption of key equipment such as ventilation, air conditioning, and dehumidification. Based on this, real-time updates of event data driven by environmental changes provide immediate information on potential environmental disturbances, helping to understand the immediate impact of external events on cleanroom air quality and enhance foresight and responsiveness to environmental changes.

[0030] During implementation, personnel monitoring relies primarily on the cleanroom's existing access control system. This system, typically deployed at cleanroom entrances and exits, is equipped with a card reader or facial recognition device to accurately record every entry and exit. Each time a person passes through the access control system, the system automatically updates the current cleanroom headcount and uploads this information to a central data acquisition module at a fixed frequency (e.g., every 10 seconds). Furthermore, the access control system provides timestamps, allowing for clear documentation of changes in headcount, facilitating subsequent analysis of their impact on air quality.

[0031] Meanwhile, door status monitoring relies on magnetic or infrared proximity sensors installed on each access door. These sensors sense whether the door is open or closed in real time and convert the status signal into a digital output. A data acquisition system regularly polls the status of all doors, ensuring that even brief openings are accurately captured, allowing identification of any impacts of external air disturbances on the cleanroom environment.

[0032] To obtain the external ambient temperature, the system uses high-precision temperature sensors independently located outside the cleanroom. These sensors are typically installed on building exterior walls or near ventilation openings, away from direct sunlight or other heat sources, to ensure the authenticity and representativeness of the measurement results. The sensors communicate with the data acquisition center via wired or wireless means, continuously transmitting external temperature data according to a preset sampling cycle. Because the cleanroom's air supply system often draws in fresh air from the outdoors, changes in external temperature directly affect the supply air temperature, thereby indirectly affecting the thermal equilibrium within the cleanroom. Therefore, real-time monitoring of external temperature not only helps understand the current environmental context but also provides basic data support for predicting possible future temperature and humidity fluctuations.

[0033] The collection of operating power for various cleanroom equipment primarily relies on smart meters, typically installed in the power supply circuits of ventilation systems, air conditioning units, dehumidifiers, and other key power equipment. These meters provide real-time measurement of electrical parameters such as voltage, current, power factor, and active power. This data is regularly collected and uploaded to a central control system via standard communication protocols such as Modbus and BACnet. Continuous monitoring of equipment operating power can identify trends in equipment status, such as sudden shutdowns, abnormal power increases, or periodic fluctuations, and thus assess their potential impact on cleanroom air quality.

[0034] Throughout the data collection process, the system must ensure that data from different devices and sensor types remains highly synchronized in time. To this end, a unified time server is typically used to calibrate all collection terminals, ensuring that time deviations between nodes are controlled within milliseconds. Furthermore, to address potential network delays, device failures, or signal loss, the system incorporates automatic retransmission mechanisms and data interpolation capabilities to ensure the integrity and continuity of the data stream. All raw data is timestamped after collection and stored in a high-performance database for subsequent use in constructing dynamic event sequences.

[0035] In the above-mentioned AI-based dynamic energy consumption management method for clean rooms, the step S3 inputs the time series of air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain short-term air quality parameter prediction results. It should be understood that since the evolution of clean room air quality is the result of the nonlinear coupling between the inertia of the historical state of the indoor environment and real-time external disturbance events, traditional linear models or simple rule engines are difficult to accurately characterize this dynamic relationship. Therefore, in order to achieve high-reliability predictions of air quality in a short period of time in the future, this application is based on deep learning technology, and by constructing an air quality AI prediction model, it is to explore the complex interactive relationship between historical time series data of air quality and real-time environmental change driving events, thereby realizing intelligent reasoning about the future trajectory of air quality. Among them, Figure 3 Flowchart of sub-step S3 of the clean room dynamic energy consumption management method based on AI according to an embodiment of the present application. Figure 3 As shown, the step S3 includes the steps of: S31, extracting the air quality time series change characteristics from the time series of the air quality parameters to obtain an air quality time series feature coding vector; S32, performing structured embedded coding on the environmental change driven event data to obtain an environmental change driven event embedded coding feature vector; S33, performing feature interactive response reasoning on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to obtain an air quality time series evolution feature interactive response reasoning coding vector; S34, performing predictive decoding on the air quality time series evolution feature interactive response reasoning coding vector to obtain the short-term air quality parameter prediction result.

[0036] Specifically, the step S31 extracts the air quality time series variation characteristics from the time series of the air quality parameters to obtain an air quality time series feature encoding vector. In a specific example of the present application, the step S31 includes: performing air quality time series feature extraction based on a forward LSTM network on the time series of the air quality parameters to obtain the air quality time series feature encoding vector. Specifically, since the evolution of clean room air quality has strong time series dependence and state memory (such as temperature changes are affected by historical heat accumulation, and PM2.5 diffusion follows the law of fluid dynamics continuity), the original discrete air quality parameter time series data cannot directly reveal the deep evolution law of air quality. Therefore, in order to explore the intrinsic dynamic characteristics of parameter changes, this solution adopts a forward LSTM (long short-term memory network) model to process the time series of the air quality parameters, so as to utilize the powerful long-term dependency modeling capability of the LSTM network to learn the evolution pattern of the air quality parameters in the time dimension. Specifically, a four-dimensional parameter sequence of temperature, humidity, PM2.5 concentration, and CO2 concentration, collected at fixed sampling intervals, is first fed into a forward LSTM network. The input vector for each time step in the LSTM network is the four-dimensional air quality parameter vector at the corresponding moment, incorporating the air quality status information at that moment. Subsequently, the LSTM network controls the flow of information through three nonlinear control units: a forget gate, an input gate, and an output gate. This control selectively retains key information from the historical state (such as the continuously rising CO2 trend) and filters out transient noise (such as occasional sensor jitter), updating the state representation for the current time step. Based on this, the network iteratively updates the hidden state vector, gradually accumulating and propagating historical time series information. Ultimately, a high-dimensional feature representation is generated that captures the historical evolution of air quality, forming a fixed-dimensional air quality time series feature encoding vector. In this way, this air quality time series feature encoding vector not only captures the continuous changing trends of air quality parameters over time but also incorporates the potential correlations and interactions between parameters, providing effective feature input for subsequent air quality prediction.

[0037] Specifically, the step S32 performs structured embedded coding on the environmental change driven event data to obtain an embedded coding feature vector of the environmental change driven event. It should be understood that the present application takes into account that the sources of the environmental change driven event data are diverse and of different types, including both numerical data (such as the operating power of indoor equipment) and discrete categorical data (such as the "open" or "closed" status of the door). Therefore, in order to convert the multi-source heterogeneous environmental change driven event data into a unified, standardized vector representation that can be understood and calculated by the AI model, the present application is based on the embedded coding technology in representation learning, and maps various types of driving events to the same high-dimensional semantic space to achieve quantitative representation of their inherent correlation. Specifically, for numerical data, such as the operating power of indoor equipment, the external ambient temperature and the number of people, based on data statistical analysis, a normalization or standardization method is used to map them to the numerical range of [0,1] to eliminate the impact of data of different magnitudes on model training. For categorical data, such as door switch status, one-hot encoding is used to convert it into a binary vector with a length equal to the total number of categories. The only position in the vector that is 1 corresponds to the index of the current category, thus achieving a numerical representation of the categorical information. Furthermore, by concatenating the numerical feature representations of various driving events and further fusing them through a fully connected neural network layer, a fixed-dimensional embedded encoding feature vector of environmental change driving events is formed. This enhances the ability to capture complex relationships between events and provides a unified data representation for subsequent air quality predictions.

[0038] Specifically, the step S33 performs feature interaction response reasoning on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to obtain the air quality time series evolution feature interaction response reasoning coding vector. Specifically, considering the complex cross-modal coupling between the historical state of indoor air quality and the emergency event (such as the door opening event will destroy the equilibrium state of the original PM2.5 concentration field, causing the concentration to rise, but the degree of destruction depends on the historical concentration gradient of indoor PM2.5). Therefore, in order to characterize the complex cross-modal interaction relationship between the historical state of indoor air quality and the environmental change driven event, the present application further simulates the causal reasoning chain of event triggering-parameter response, and performs feature interaction response reasoning on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to generate a comprehensive feature representation that includes the interactive impact of the air quality historical state and real-time events, and obtains the air quality time series evolution feature interaction response reasoning coding vector. Among them, Figure 4 Flowchart of sub-step S33 of the clean room dynamic energy consumption management method based on AI according to an embodiment of the present application. Figure 4As shown, the step S33 includes the steps of: S331, performing fine-grained feature interaction response encoding on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to obtain a sequence distribution of the environmental drive-air quality time series feature local interaction response feature coding vector; S332, performing global interaction reasoning on the sequence distribution of the environmental drive-air quality time series feature local interaction response feature coding vector to obtain the air quality time series evolution feature interaction response reasoning coding vector.

[0039] More specifically, in step S331, fine-grained feature interaction response coding is performed on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to obtain a sequence distribution of the environmental drive-air quality time series feature local interaction response feature coding vector. Figure 5 Flowchart of sub-step S331 of the clean room dynamic energy consumption management method based on AI according to an embodiment of the present application. Figure 5 As shown, the step S331 includes the steps of: S3311, performing dual-channel interactive fusion based on tanh function asymmetric interaction and sigmoid function gated interaction on the air quality time series feature coding vector and the environmental change driven event embedded coding feature vector to obtain the environmental drive-air quality time series feature global fine-grained interaction response matrix; S3312, performing matrix decomposition on the environmental drive-air quality time series feature global fine-grained interaction response matrix to obtain the sequence distribution of the environmental drive-air quality time series feature local interaction response feature coding vector.

[0040] In a specific example of the present application, step S3311 is expressed as follows: in, represents the hyperbolic tangent function, Represents the air quality time series feature encoding vector, Indicates the embedding encoding feature vector of the environmental change driving event, represents the transpose of a vector, represents the sigmoid activation function, is the dynamic weight matrix, is the projection matrix, Indicates point multiplication by position, It means adding by position. represents matrix multiplication, is the hyperbolic tangent asymmetric interaction, For element-level gated interactions, Represents the global fine-grained interaction response matrix of environmental driving and air quality time series characteristics.

[0041] Specifically, the asymmetric interaction of the tanh function characterizes the degree of asymmetric influence between the air quality time series feature encoding vector and the environmental change-driven event embedding encoding feature vector, while the gated interaction of the sigmoid function enables selective activation of key interaction features. The combination of the two can deeply explore the potential connections between the two types of feature vectors from different perspectives. This dual-channel interactive fusion approach accurately captures subtle semantic resonances and differences between features while ensuring comprehensive interaction. The resulting global, fine-grained interaction response matrix of environmental-driven-air quality time series features has extremely high information density, deeply characterizing the complex coupling relationship between the air quality time series feature encoding vector and the environmental change-driven event embedding encoding feature vector, providing richer and more accurate feature interaction information for subsequent reasoning processes.

[0042] In particular, considering that when calculating the global fine-grained interaction response matrix of the environmental driving-air quality time series characteristics, due to the fine-grained correlation matrix Dynamic weight matrices were used and the projection matrix Modulation in different directions, and the function and Having different orders of exponential functions based on natural constants, while ensuring full coverage of global interactions, may cause the subtle semantic representations at fine scales to differentiate along different coupling channels, thereby affecting the fusion effect of the two parts. Therefore, in order to improve the fusion effect of subtle semantic representations at fine granularity. In a preferred example of the present application, the step S3312 includes: first, performing a dual-channel interaction feature equivalence correction on the global fine-grained interaction response matrix of the environmental drive-air quality time series features to obtain an optimized global fine-grained interaction response matrix of the environmental drive-air quality time series features.

[0043] Specifically, let ,and , first obtain the transformation matrix by projective equivariant sequential coupling verification operation ,satisfy: in, Represents the hyperbolic tangent asymmetric interaction matrix of environmental driving-air quality time series characteristics, represents the environmental driving-air quality temporal feature gated interaction matrix, Represents the transformation matrix.

[0044] That is, it is believed that there is a substantial interaction between different projection parts, so that the transformation matrix Generate corresponding expressions based on different interaction sequences.

[0045] Then, we can get the transformation matrix Perform eigenstate degenerate reconstruction, that is: in, represents the inverse matrix, express norm, represents the equalization correction matrix.

[0046] That is, the transformation matrix and its inverse matrix As the group representation of the generator representation and the inverse element representation, respectively, through the matrix The norm is used to degenerate the state space, that is, to reduce the group theory representation of the state space, so as to obtain the interaction relationship representation based on the ground state. In this way, the unidirectionality of the generator representation and the inverse representation can be realized based on the differential representation of the evolution between states.

[0047] Finally, the equalization correction matrix To calibrate and ,Right now , Then fusion ,in represents the hyperbolic tangent asymmetric interaction matrix of the corrected environmental driving-air quality time series characteristics, represents the gated interaction matrix of environmental driving-air quality time series characteristics after correction, It represents the optimization of the global fine-grained interaction response matrix of environmental driving and air quality time series characteristics, thereby improving the fusion completeness of the original global fine-grained interaction response matrix of environmental driving and air quality time series characteristics.

[0048] Then, the optimized global fine-grained interaction response matrix of the environmental driving-air quality time series feature is decomposed into column vectors to obtain the sequence distribution of the local interaction response feature encoding vector of the environmental driving-air quality time series feature, which is expressed as follows: in, represents the matrix decomposition function, 、 、 and They represent the first, second, and third order in the sequence distribution of the local interactive response feature encoding vector of the environmental driving-air quality temporal characteristics. and The local interaction response feature encoding vector of environmental driving-air quality time series features.

[0049] Specifically, by decomposing the global fine-grained interaction response matrix of the optimized environmental driver-air quality temporal features by columns, the global interaction information is deconstructed into local information flows with temporal dynamics, enabling each generated feature encoding vector of the local interaction response of the environmental driver-air quality temporal features to correspond to a phased semantic response from a specific interaction perspective. This matrix decomposition method breaks the static constraints of the two-dimensional matrix, transforming the global interaction information into a one-dimensional local response sequence. This preserves the fine-grained feature correlation information of the optimized global fine-grained interaction response matrix of the environmental driver-air quality temporal features, and provides a feature expression that is closer to cognitive laws for intelligent reasoning about the temporal evolution of air quality. This in turn improves the depth of understanding and reasoning accuracy of the prediction model regarding the dynamic relationship between environmental change-driven events and air quality.

[0050] More specifically, in a specific example of the present application, step S332 includes: inputting the sequence distribution of the local interactive response feature encoding vector of the environmental drive-air quality time series feature into the response inference generation engine based on the forward LSTM model to obtain the interactive response inference encoding vector of the air quality time series evolution feature, which is expressed as follows: in, represents the forward LSTM model, Represents the interactive response inference encoding vector of the time series evolution characteristics of air quality.

[0051] That is, by leveraging the forward LSTM model's ability to model the temporal distribution of sequences, the feature encoding vectors of the local interactive responses of environmental driver-air quality temporal features obtained through static decomposition are transformed into a holistic representation of dynamic associations. Specifically, by sequentially processing each local interactive response feature encoding vector of environmental driver-air quality temporal features, the forward LSTM model integrates current input with historical contextual information while iteratively updating its internal state, ultimately generating an inference encoding vector of the interactive response of the temporal evolution of air quality features that not only contains the independent information of each local response but also encodes the logic of its temporal evolution. This provides a deep feature representation that integrates the dynamic associations of the time dimension for subsequent air quality prediction.

[0052] Specifically, step S34 involves predictively decoding the air quality temporal evolution feature interactive response inference encoding vector to obtain the short-term air quality parameter prediction result. Specifically, to decode the complex feature information in the air quality temporal evolution feature interactive response inference encoding vector into specific, multi-time-step air quality parameter prediction values, this application constructs an RNN (recurrent neural network)-based air quality prediction decoding layer based on sequence generation technology. The air quality temporal evolution feature interactive response inference encoding vector serves as the initial hidden state of the decoding layer, and an autoregressive generation is performed with a set prediction time range. At each time step, the air quality parameter value for the next time step is predicted based on the current hidden state and the predicted context information. For example, at the first time step, the air quality parameter for the first future time step (t+1) is predicted based on the initial state. Next, the predicted parameter for time t+1 is used as the new context information and combined with the updated hidden state to predict the air quality parameter for the second future time step (t+2). This process continues until all air quality parameter values within the set time range are predicted, providing forward-looking decision support for subsequent air quality management and regulation.

[0053] In the above-mentioned AI-based dynamic energy consumption management method for clean rooms, the step S4 generates an air quality adjustment instruction based on the comparison between the short-term air quality parameter prediction result and the preset air quality standard range. It should be understood that the short-term air quality parameter prediction result represents the possible state of indoor air quality in the future under the influence of the current driving event. By comparing and analyzing the prediction result with the preset air quality standard range, it can be determined whether the future indoor air quality meets the standard, and then the corresponding air quality adjustment can be performed. Among them, Figure 6 Flowchart of sub-step S4 of the clean room dynamic energy consumption management method based on AI according to an embodiment of the present application. Figure 6 As shown, the step S4 includes the following steps: S41, determining whether each air quality parameter in the short-time air quality parameter prediction result is within the preset air quality standard range; if so, generating an air quality adjustment instruction to keep the current operating parameters of the HVAC system unchanged; S42, if there is at least one time point in the short-time air quality parameter prediction result where the air quality parameter is not within the preset air quality standard range, recording the over-limit direction and maximum over-limit amplitude of the short-time air quality parameter prediction result, and calculating the feedforward compensation target parameter at the current time point based on the over-limit direction and the maximum over-limit amplitude as the target setting value of the PID controller to generate an air quality adjustment instruction.

[0054] Specifically, step S41 determines whether each air quality parameter in the short-term air quality parameter prediction results is within the preset air quality standard range. If so, an air quality adjustment instruction is generated to maintain the current operating parameters of the HVAC system unchanged. That is, the short-term air quality parameter prediction results are compared point by point with the preset cleanliness standard range (e.g., temperature 22±1°C, PM2.5 <10μg / m³). If the prediction results show that all parameters are within the standard range at all future time points, the system determines that the current HVAC operating state is sufficient to cope with future environmental changes and generates an instruction to maintain the current operating parameters unchanged, thereby avoiding unnecessary energy consumption while meeting cleanliness requirements.

[0055] Specifically, in step S42, if the air quality parameter at at least one time point in the short-term air quality parameter prediction result is not within the preset air quality standard range, the exceeding direction and the maximum exceeding amplitude of the short-term air quality parameter prediction result are recorded, and the feedforward compensation target parameter at the current time point is calculated based on the exceeding direction and the maximum exceeding amplitude as the target setting value of the PID controller to generate an air quality adjustment instruction. That is, if it is predicted that a certain parameter at at least one time point in the future will exceed the standard range, the exceeding direction (whether it is too high or too low) and the maximum exceeding amplitude are further recorded. Then, the feedforward compensation target parameter is calculated according to the preset control algorithm. For example, if it is predicted that the temperature will exceed the upper limit of 0.8°C in 10 minutes, the exceeding amplitude of 0.8°C is multiplied by the preset weight. ( ), subtract the current temperature value of 22.5℃ from 0.8℃ to obtain the feedforward compensation target parameter at the current time point. Then, the feedforward compensation target parameter at the current time point is used as the target set value of the PID controller. After receiving the updated target set value, the PID controller combines the real-time air quality monitoring data and dynamically adjusts the operating parameters of the HVAC system, such as cooling power and air supply volume, based on the deviation between the target set value and the real-time monitoring data, to intervene in advance and guide the indoor air quality to a state that meets the standard. In this way, not only is the hysteresis of traditional feedback control eliminated and the actual exceeding of the air quality standard avoided, but the adjustment action of the HVAC system is also made smooth and continuous, eliminating drastic fluctuations and peaks in energy consumption. This maximizes energy utilization while ensuring air quality, thereby improving the overall comfort and energy efficiency of the indoor environment.

[0056] In summary, the AI-based dynamic energy consumption management method for clean rooms based on the embodiment of the present application is explained. It introduces a deep learning algorithm to conduct in-depth interactive analysis of the time series data of clean room air quality parameters and environmental change driving event data such as the number of indoor people, external ambient temperature, door switch status, and equipment power, so as to realize intelligent reasoning of the dynamic evolution process of clean room air quality, thereby predicting the air quality changes in the short term in the future under the current driving event, and then, by comparing the short-term air quality parameter prediction results with the preset air quality standard range, the HVAC system operating parameters of the clean room are feedforward adjusted according to the difference between the two. This method can proactively foresee environmental changes and perform smooth pre-adjustment in advance to maintain the air quality of the clean room within the preset standard range, while significantly reducing unnecessary energy waste and improving environmental stability.

[0057] Furthermore, an AI-based clean room dynamic energy consumption management system is also provided.

[0058] Figure 7 FIG is a block diagram of an AI-based clean room dynamic energy management system according to an embodiment of the present application. Figure 7 As shown, according to an embodiment of the present application, the AI-based clean room dynamic energy consumption management system 100 includes: an air quality parameter acquisition module 110, which is used to obtain the time series of the air quality parameters of the clean room within a preset time window including the current time point, and the air quality parameters include temperature, humidity, PM2.5 concentration and carbon dioxide concentration; a driving event data acquisition module 120, which is used to obtain the environmental change driving event data of the clean room in real time, and the environmental change driving event data includes the current number of people in the clean room, the door switch status, the external ambient temperature, and the operating power of the indoor equipment; an air quality prediction module 130, which is used to input the time series of the air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain a short-term air quality parameter prediction result; an air quality adjustment instruction generation module 140, which is used to generate an air quality adjustment instruction based on a comparison between the short-term air quality parameter prediction result and the preset air quality standard range.

[0059] Here, those skilled in the art will understand that the specific operations of each module in the above-mentioned AI-based clean room dynamic energy consumption management system have been referred to above. Figures 1 to 6 The description of the AI-based clean room dynamic energy consumption management method has been introduced in detail, and therefore, its repeated description will be omitted.

[0060] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details of the above embodiments are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0061] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0062] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be encompassed therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0063] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0064] Finally, it should be noted that the above description has been provided for the purpose of illustration and description. In addition, the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the technical solutions may be modified or replaced with equivalents with reference to the preferred embodiments, they do not depart from the spirit and scope of the technical solutions of the present invention.

Claims

1. A clean room dynamic energy consumption management method based on AI, characterized in that: include: Obtain a time series of air quality parameters of the clean room within a preset time window including the current time point, wherein the air quality parameters include temperature, humidity, PM2.5 concentration, and carbon dioxide concentration; Acquire environmental change-driven event data of the clean room in real time, wherein the environmental change-driven event data includes the current number of people in the clean room, the door switch status, the external ambient temperature, and the operating power of the indoor equipment; Inputting the time series of air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain a short-term air quality parameter prediction result; An air quality adjustment instruction is generated based on a comparison between the short-term air quality parameter prediction result and a preset air quality standard range.

2. The AI-based clean room dynamic energy consumption management method according to claim 1 is characterized in that: Generating an air quality adjustment instruction based on a comparison between the short-term air quality parameter prediction result and a preset air quality standard range includes: determining whether each air quality parameter in the short-term air quality parameter prediction result is within the preset air quality standard range, and if so, generating an air quality adjustment instruction to keep the current operating parameters of the HVAC system unchanged; If the air quality parameters of at least one time point in the short-term air quality parameter prediction result are outside the preset air quality standard range, the exceeding direction and maximum exceeding amplitude of the short-term air quality parameter prediction result are recorded, and the feedforward compensation target parameter at the current time point is calculated based on the exceeding direction and the maximum exceeding amplitude as the target setting value of the PID controller to generate an air quality adjustment instruction.

3. The AI-based clean room dynamic energy consumption management method according to claim 1, characterized in that: Inputting the time series of the air quality parameters within the preset time window including the current time point and the environmental change driving event data into the air quality AI prediction model to obtain a short-term air quality parameter prediction result, including: Extracting air quality time series variation characteristics from the time series of the air quality parameters to obtain an air quality time series feature encoding vector; Performing structured embedding coding on the environmental change driven event data to obtain an environmental change driven event embedding coding feature vector; Performing feature interaction response reasoning on the air quality time series feature coding vector and the environmental change driving event embedded coding feature vector to obtain an air quality time series evolution feature interaction response reasoning coding vector; The air quality time series evolution feature interactive response inference coding vector is predicted and decoded to obtain the short-term air quality parameter prediction result.

4. The AI-based clean room dynamic energy consumption management method according to claim 3 is characterized in that: Extracting air quality temporal variation characteristics from the time series of the air quality parameter to obtain an air quality temporal feature encoding vector includes: Air quality time series feature extraction based on a forward LSTM network is performed on the time series of the air quality parameter to obtain the air quality time series feature encoding vector.

5. The AI-based clean room dynamic energy consumption management method according to claim 4 is characterized in that: Performing feature interaction response reasoning on the air quality time series feature coding vector and the environmental change driving event embedded coding feature vector to obtain an air quality time series evolution feature interaction response reasoning coding vector, including: Performing fine-grained feature interaction response coding on the air quality time series feature coding vector and the environmental change driving event embedded coding feature vector to obtain a sequence distribution of the environmental driving-air quality time series feature local interaction response feature coding vector; Global interactive reasoning is performed on the sequence distribution of the local interactive response feature coding vector of the environmental drive-air quality time series feature to obtain the interactive response reasoning coding vector of the air quality time series evolution feature.

6. The AI-based clean room dynamic energy consumption management method according to claim 5, characterized in that: Fine-grained feature interaction response encoding is performed on the air quality time series feature encoding vector and the environmental change driving event embedded encoding feature vector to obtain a sequence distribution of the environmental driving-air quality time series feature local interaction response feature encoding vector, including: Performing dual-channel interactive fusion based on tanh function asymmetric interaction and sigmoid function gated interaction on the air quality time series feature encoding vector and the environmental change driving event embedded encoding feature vector to obtain a global fine-grained interactive response matrix of environmental driving-air quality time series features; The global fine-grained interaction response matrix of the environmental driving-air quality time series characteristics is subjected to matrix decomposition to obtain a sequence distribution of the local interaction response feature encoding vectors of the environmental driving-air quality time series characteristics.

7. The AI-based clean room dynamic energy consumption management method according to claim 6, characterized in that: Performing matrix decomposition on the global fine-grained interaction response matrix of the environmental driving-air quality time series feature to obtain a sequence distribution of the local interaction response feature encoding vector of the environmental driving-air quality time series feature, including: Performing dual-channel interaction feature equivalence correction on the global fine-grained interaction response matrix of the environmental drive-air quality time series characteristics to obtain an optimized global fine-grained interaction response matrix of the environmental drive-air quality time series characteristics; The optimized environmental driving-air quality time series feature global fine-grained interaction response matrix is matrix decomposed according to column vectors to obtain the sequence distribution of the environmental driving-air quality time series feature local interaction response feature encoding vectors.

8. The AI-based clean room dynamic energy consumption management method according to claim 7, characterized in that: Performing global interactive reasoning on the sequence distribution of the local interactive response feature coding vector of the environmental drive-air quality time series feature to obtain the interactive response reasoning coding vector of the air quality time series evolution feature includes: The sequence distribution of the local interactive response feature encoding vector of the environmental drive-air quality time series feature is input into the response inference generation engine based on the forward LSTM model to obtain the interactive response inference encoding vector of the air quality time series evolution feature.

9. The AI-based clean room dynamic energy consumption management method according to claim 5, characterized in that: Predictively decoding the air quality time series evolution feature interactive response inference coding vector to obtain the short-term air quality parameter prediction result includes: The air quality time series evolution feature interactive response inference coding vector is input into the RNN-based air quality prediction decoding layer to obtain the short-term air quality parameter prediction result.

10. An AI-based clean room dynamic energy consumption management system, characterized in that: include: An air quality parameter acquisition module is used to obtain a time series of air quality parameters of the clean room within a preset time window including the current time point, wherein the air quality parameters include temperature, humidity, PM2.5 concentration and carbon dioxide concentration; A drive event data acquisition module is used to acquire the environmental change drive event data of the clean room in real time, wherein the environmental change drive event data includes the current number of people in the clean room, the door switch status, the external environment temperature, and the operating power of the indoor equipment; An air quality prediction module is configured to input the time series of air quality parameters within a preset time window including the current time point and the environmental change driving event data into an air quality AI prediction model to obtain a short-term air quality parameter prediction result; The air quality adjustment instruction generating module is used to generate an air quality adjustment instruction based on the comparison between the short-term air quality parameter prediction result and the preset air quality standard range.

Images