An intelligent regulation and control method and system for an air conditioning system of an industrial clean workshop based on production environment parameter prediction

Abstract

The application discloses an intelligent regulation and control method and system for an air conditioning system of an industrial clean workshop based on production environment parameter prediction, comprising a construction operation management database, a random forest regression model and a joint simulation platform; based on current weather variable factors, the random forest regression model predicts operation management data of the air conditioning system; a reference working condition point close to the current weather variable factors is selected in the operation management database; historical simulation data and historical operation management data corresponding to the reference working condition point are taken as initial values of the platform to perform simulation, and risk point data and simulation operation management data are obtained; recommended operation management data under the current weather variable factors is calculated by using the simulation operation management data and the predicted operation management data; and the air conditioning system is adjusted based on the risk point data and the recommended operation management data. The application determines the control strategy of the air conditioning system by using the predicted operation management data of the model and the operation management data simulated by the joint simulation platform, provides accurate recommended operation management data under the current weather variable factors, and improves the precision and stability of the air conditioning system control.

Description

Technical Field

[0001] This invention relates to the field of HVAC system optimization control technology, and in particular to an intelligent control method and system for industrial cleanroom air conditioning systems based on the prediction of production environment parameters. Background Technology

[0002] In the production environment of high-tech factories, most existing air conditioning systems rely on feedback regulation after indoor environmental parameters reach preset thresholds. Methods such as Proportional-Integral-Derivative (PID) control or Programmable Logic Controllers (PLCs) primarily depend on comparing actual parameters (such as temperature and humidity) monitored by indoor environmental sensors with preset values. When actual parameters deviate from the preset values, the operating status of the air conditioning system is adjusted, such as regulating the airflow and chilled water flow, to restore indoor environmental stability. However, the regulatory effect of feedback control lags behind the effect of disturbances, exhibiting a time lag. This causes fluctuations in indoor environmental parameters within a short period, affecting the stability of the production environment.

[0003] Existing model predictive control (MMDC) combines feedback control and feedforward control. Feedforward control is achieved by predicting disturbances using a predictive model, and feedback control is achieved by correcting the predicted values ​​based on the error between the predicted and measured values. The corrected predicted values ​​are then input into a rolling optimization module, where an optimization algorithm solves the objective function. The first control vector in the result becomes the controller's output signal for the next time step, achieving rolling optimization. However, the quality and completeness of the data are crucial to the accuracy of model prediction. Current issues such as the lack of large datasets for algorithm training and significant differences between training data and subsequent real data characteristics negatively impact the accuracy of MMDC. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide an intelligent control method and system for air conditioning systems in industrial cleanrooms based on the prediction of production environment parameters, so as to improve the accuracy and stability of air conditioning system control and reduce control lag.

[0005] Technical Solution: To achieve the above objectives, the present invention provides an intelligent control method for an industrial cleanroom air conditioning system based on the prediction of production environment parameters, comprising the following steps:

[0006] S1: Build an operation and maintenance management database, including multiple sets of historical data. Each set of historical data includes weather variable factors with historical timestamps, as well as historical predicted operation and maintenance data, historical simulation data, and historical operation and maintenance data corresponding to the weather variable factors.

[0007] S2: Establish a random forest regression model between weather variables and historical operation and maintenance data;

[0008] S3: Obtain current weather variables, use a random forest regression model to make predictions, obtain predicted operation and maintenance data, and store it in the operation and maintenance management database;

[0009] S4: Based on the current weather variable factors, perform sensitivity analysis on the weather variable factors in the operation and maintenance management database, and use the operating point with the most similar sensitive conditions as the reference operating point near the current weather variable factor;

[0010] S5: Build a joint simulation platform, including simulation models of the factory building complex, air conditioning system, and factory production environment;

[0011] S6: Use the historical simulation data and historical operation and maintenance data corresponding to the nearby reference working condition point as the initial values ​​of the joint simulation platform to perform simulation, obtain the risk point data and simulation operation and maintenance data under the current weather variable factors, and store them in the operation and maintenance management database.

[0012] S7: Weight the simulated operation and maintenance data and the predicted operation and maintenance data to obtain the recommended operation and maintenance data under the current weather variable factors, and store it in the operation and maintenance management database;

[0013] S8: Upload monitoring risk point data and recommended operation and maintenance data to the terminal to enable automatic or manual adjustment of the air conditioning system.

[0014] Among them, the weather variables mentioned in S1 include air pressure, weather conditions, precipitation probability, precipitation amount, temperature, perceived temperature, relative humidity, wind direction, wind speed, gusts, and cloud cover; the operating point refers to the air conditioning system operation and maintenance data under specific weather variable conditions; the historical simulation data includes historical risk point data and historical simulation data; the air conditioning system operation and maintenance data includes, but is not limited to, cooling water temperature, cooling water flow rate, chilled water temperature, chilled water flow rate, supply air temperature, supply air volume, return air volume, fresh air volume, cleanroom temperature, cleanroom humidity, cleanroom pressure and particle count, water valve opening, air valve opening, equipment start-up and shutdown records, fault records, alarm information, and anomaly records.

[0015] Specifically, S2 describes establishing a random forest regression model between weather variables and historical operation and maintenance data, including:

[0016] S201. Data preprocessing: Quantify, clean, and standardize the weather variable factors and historical operation and maintenance data coding methods in each set of historical data.

[0017] S202. Establish a random forest regression model: Use the preprocessed weather variables in each set of data as feature variables X, and the historical operation and maintenance data as target variables Y, and train the random forest regression model to obtain the trained random forest regression model.

[0018] The method for obtaining the reference operating point through sensitivity analysis in S4 is as follows:

[0019] S401. Use the trained random forest model to calculate the feature importance score of weather variable factors in each group of historical data, and calculate the weight of weather variable factors based on feature importance.

[0020] S402. Match the current weather variable factors with the weather variable factors in each set of historical data. For each historical data point, use weights to calculate the weighted Euclidean distance between the current weather variable factors and the historical weather variable factors, and calculate the weighted Euclidean distance between the current expected operating point and the historical operating point.

[0021] S403. Traverse the historical data, calculate the comprehensive distance of each set of historical data, and select the working point corresponding to the historical data with the smallest comprehensive distance as the reference working point for the current weather variable factor.

[0022] The co-simulation platform described in S5 employs CAE (Computer-Aided Engineering) technology and the Modelica language. The factory building complex simulation model serves as the top-level model, providing dynamic data of the external environment as input conditions for the air conditioning system simulation model and the factory production environment simulation model. The air conditioning system simulation model and the factory production environment simulation model receive output data from the factory building complex simulation model as input conditions. Simulation results are exchanged between the air conditioning system simulation model and the factory production environment simulation model as input conditions for each other, forming a closed-loop co-simulation platform.

[0023] Specifically, S7 describes weighting the simulated operation and maintenance data and the predicted operation and maintenance data to obtain recommended operation and maintenance data under the current weather variable factors, including:

[0024] Calculate the weighting coefficients for simulated operation and maintenance data and predicted operation and maintenance data. {CAE} weight {reg} Then, the simulation operation and maintenance data described in S6 is predicted. {CAE} The prediction of operation and maintenance data from the random forest regression model described in S3. {reg} Weighted calculations are performed to obtain the recommended operation and maintenance data (prediction) under the current weather variable factors. {final} :

[0025] prediction {final}=weight {reg} ×prediction {reg} +weight {CAE} ×prediction {CAE} .

[0026] Among them, the weight coefficients for simulated operation and maintenance data and predicted operation and maintenance data are... {CAE} weight {reg} The calculation method is as follows:

[0027] Calculate the historical error between the historical predicted operation and maintenance data and the historical operation and maintenance data corresponding to each historical weather variable factor in the operation and maintenance management database, and set the error of the predicted operation and maintenance data to follow a certain probability distribution;

[0028] Calculate the historical error between historical simulation data and historical operation and maintenance data corresponding to each historical weather variable factor in the operation and maintenance management database, and set the error of the simulated operation and maintenance data to follow a certain probability distribution;

[0029] Calculate the probability density of predicted operation and maintenance data errors based on probability distribution (pdf) {reg} Probability density of simulation operation and maintenance data errors (pdf) {CAE} sum:

[0030] pdf {total} =pdf {reg} +pdf {CAE} ,

[0031] The weighting coefficients for the predicted operation and maintenance data corresponding to the random forest regression model are:

[0032]

[0033] The weighting coefficients for the simulation operation and maintenance data corresponding to the co-simulation platform are:

[0034]

[0035] The present invention discloses an intelligent control system for air conditioning systems in industrial cleanrooms based on the prediction of production environment parameters, comprising:

[0036] Database module: Build an operation and maintenance management database, including multiple sets of historical data. Each set of historical data includes weather variable factors with historical timestamps, as well as historical predicted operation and maintenance data, historical simulation data, and historical operation and maintenance data corresponding to the weather variable factors.

[0037] Predictive model building module: Establishes a random forest regression model between weather variables and historical operation and maintenance data;

[0038] Predictive Operation and Maintenance Data Generation Module: Obtains current weather variables and uses a random forest regression model to make predictions, thereby obtaining predictive operation and maintenance data;

[0039] Reference operating condition point acquisition module: Based on the current weather variable factors, perform sensitivity analysis on the weather variable factors in the operation and maintenance management database, and use the operating condition point with the most similar sensitive conditions as the reference operating condition point near the current weather variable factor;

[0040] Co-simulation platform construction module: Builds a co-simulation platform, including simulation models of the factory building complex, air conditioning system, and factory production environment;

[0041] Simulation data generation module: It uses historical simulation data and historical operation and maintenance data corresponding to nearby reference working conditions as the initial values ​​of the joint simulation platform to simulate and obtain risk point data and simulation operation and maintenance data under the current weather variable factors, and stores them in the operation and maintenance management database.

[0042] Recommended Operation and Maintenance Data Generation Module: This module performs weighted calculations on simulated and predicted operation and maintenance data to obtain recommended operation and maintenance data under the current weather variable factors.

[0043] Data upload module: Uploads monitoring risk point data and recommended operation and maintenance data to the terminal to enable automatic or manual adjustment of the air conditioning system.

[0044] Beneficial Effects: This invention has the following advantages: 1. By establishing a random forest regression model, this invention can predict the operation and maintenance needs of the air conditioning system based on the current weather variable factors, thus predicting the operation and maintenance needs of the air conditioning system before fluctuations in production environment parameters, reducing the control lag of the air conditioning system; 2. By building a co-simulation platform, model predictive control can be achieved on a small dataset, and early warning of production environment risks can be provided, providing a scientific basis for operation and maintenance decisions; 3. By using a small dataset to initially achieve predictive control of the random forest regression model, and by weighting the simulation results of the co-simulation platform and the prediction results of the random forest regression model, the control strategy can be determined, providing accurate recommended operation and maintenance data under the current weather variable factors, improving the accuracy and stability of air conditioning system control. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating the method. Detailed Implementation

[0046] The technical solution of the present invention will be described in detail below with reference to the embodiments and accompanying drawings.

[0047] like Figure 1 As shown, the intelligent control method for industrial cleanroom air conditioning systems based on production environment parameter prediction, as described in this invention, includes the following steps:

[0048] I. Establish an operation and maintenance management database.

[0049] The database includes multiple sets of historical data. Each set of historical data includes weather variables with historical timestamps, as well as corresponding historical predicted operation and maintenance data, historical simulation data, and historical operation and maintenance data. Weather variables include air pressure, weather conditions, precipitation probability, precipitation amount, temperature, perceived temperature, relative humidity, wind direction, wind speed, gusts, and cloud cover. Operating points represent the operation and maintenance settings of the air conditioning system under specific weather variable conditions. The historical simulation data includes historical risk point data and historical simulation data. The air conditioning system operation and maintenance data includes cooling water temperature, cooling water flow rate, chilled water temperature, chilled water flow rate, supply air temperature, supply air volume, return air volume, fresh air volume, cleanroom temperature, cleanroom humidity, cleanroom pressure and particle count, water valve opening, air valve opening, equipment start / stop records, fault records, alarm information, and anomaly records.

[0050] 2. Establish a random forest regression model between weather variables and historical operation and maintenance data.

[0051] 1. Data preprocessing: Quantify the weather variables and historical operation and maintenance data coding methods in each set of historical data (e.g., sunny is coded as 4, cloudy as 3, overcast as 2, and rain as 1), and perform cleaning and standardization.

[0052] Data cleaning is used to check for missing values. For missing values, different methods are used to handle them depending on the characteristics of the data. For example, for numerical data (such as temperature, humidity, water volume, wind volume, etc.), the median or mean is used to fill in the missing values; for categorical data (such as wind direction, weather conditions), the most common category for the three days before and after is used to fill in the missing values; for outliers, methods such as drawing box plots of the data are used to identify them. If outliers are found to deviate significantly from the normal range, they are considered for deletion or correction.

[0053] Data standardization is the process of standardizing numerical features such as "weather conditions (quantified)," "precipitation (mm)," "wind force (level)," "temperature (°C)," and "humidity (%)" by using the following formula to calculate the mean and standard deviation, so that the mean is close to 0 and the standard deviation is close to 1.

[0054]

[0055] Where x is the weather variable factor characteristic value, μ is the mean of the characteristic, σ is the standard deviation, and Z is the standardized value.

[0056] 2. Establish a random forest regression model: Use preprocessed weather variables from each set of historical data as feature variables X (e.g., quantified "weather conditions", "precipitation (mm)", "wind force (level)", "wind direction", "temperature (°C)", "humidity (%)"), and historical operation and maintenance data as target variables Y (e.g., "cooling water temperature (°C)", "chilled water temperature (°C)", "supply air temperature (°C)" for each unit, and "supply air volume (m³)" for each workshop). 3 The random forest regression model was trained using "risk point data" ( / h) and "risk point data" in the production workshop.

[0057] 3. Real-time acquisition of current weather variables (such as "weather conditions", "precipitation (mm)", "wind force (level)", "wind direction", "temperature (°C)", "humidity (%)"), preprocessing, and then using a trained random forest regression model to make predictions and obtain predicted operation and maintenance data.

[0058] IV. Based on current weather variables, conduct sensitivity analysis on the weather variables in the operation and maintenance management database. Use the operating point with the most similar sensitive conditions as the reference operating point for the current weather variables. Specifically:

[0059] 1. Feature Importance Assessment and Weight Calculation: The trained random forest model is used to calculate feature importance, obtaining an importance score for each feature. For a model with n features, the importance scores are as follows:

[0060] importance scores =[score1,score2,...,scoren];

[0061] The weights are as follows:

[0062]

[0063] 2. Calculate the weighted Euclidean distance: Match the current weather variable factors with the weather variable factors in the historical data. For each historical data point, calculate the weighted Euclidean distance between the current weather variable factor and the historical weather variable factor using weights. At the same time, calculate the weighted Euclidean distance between the current expected operating point and the historical operating point.

[0064] Current weather variables are represented as follows:

[0065] current {weather} =[s,p,w,d,t,h],

[0066] Where s, p, w, d, t, h represent the pre-processed weather conditions, precipitation, wind force, wind direction, temperature, and humidity, respectively;

[0067] Historical weather variables are:

[0068] hsitory {weather} =[s h ,p h ,w h ,d h ,t h ,h h ],

[0069] Among them, s h ,p h ,w h ,d h ,t h ,h h These represent historical data for weather conditions, precipitation, wind force, wind direction, temperature, and humidity, respectively.

[0070] The weights of each factor are:

[0071] weights = [w s ,w p ,w w ,w d ,w t ,w h ],

[0072] Among them, w s ,w p ,w w ,w d ,w t ,w h The weights of corresponding weather conditions, precipitation, wind force, wind direction, temperature, and humidity;

[0073] The formula for calculating the weighted Euclidean distance between current weather variables and historical weather variables is:

[0074] distance = np.sqrt(w s ×(ss h ) 2 +w p ×(pp h ) 2 +w w ×(ww h ) 2 +w d ×(dd h ) 2 +w t ×(tt h ) 2 +w h ×(hh h ) 2 .

[0075] The current expected operating point data is represented as follows:

[0076] current {condition} =[Q_c,T_c,Q_i,T_i,V_a,T_a],

[0077] Where Q_c, T_c, Qi, T_i, V_a, and T_a represent the chilled water flow rate, chilled water temperature, cooling water flow rate, cooling water temperature, air volume, and air temperature of each unit at the current desired operating point, respectively.

[0078] Historical operating condition data is represented as follows:

[0079] history {condition} =[Q_c h ,T_c h ,Q_i h ,T_i h ,V_a h ,T_a h ],

[0080] Among them, Q_c h ,T_c h ,Q_i h ,T_i h ,V_a h ,T_a h These represent the historical data for chilled water flow rate, chilled water temperature, cooling water flow rate, cooling water temperature, air volume, and air temperature for each unit.

[0081] The weights of each factor are expressed as follows:

[0082] weights {condition} =[w Q_c ,w T_c ,w Q_i ,w T_i ,w V_a ,w T_a ]

[0083] Among them, w Q_c ,w T_c ,w Q_i ,w T_i ,w V_a ,w T_a The weights of chilled water flow rate, chilled water temperature, cooling water flow rate, cooling water temperature, air volume, and air temperature for each unit.

[0084] The formula for calculating the weighted Euclidean distance between the current expected operating point and the historical operating point is as follows:

[0085] distance {condition} =np.sqrt(wQ_c ×(Q_c-Q_c h ) 2 +w T_c ×(T_c-T_c h ) 2 +w Q_i ×(Q_i-Q_i h ) 2 +w T_i ×(T_i-T_i h ) 2 +w V_a ×(V_a-V_a h ) 2 +w T_a ×(T_a-T_a h ) 2 .

[0086] 3. Traverse the historical data and calculate the comprehensive distance for each set of historical data, which is the sum of the weighted Euclidean distance between the current weather variable factor and the historical weather variable factor and the weighted Euclidean distance between the current expected operating point and the historical operating point. Select the operating point corresponding to the historical data with the smallest comprehensive distance as the reference operating point near the current weather variable factor.

[0087] V. Establish a joint simulation platform, including simulation models of the factory building complex, air conditioning system, and factory production environment.

[0088] Factory Building Complex Simulation Model: Geometric models are created for all buildings within the factory area, using simulations that incorporate physical fields such as flow, heat, and pollutants. Considering different wind directions, upstream pollution sources (air pollutants from surrounding factories such as power plants and electronics factories), and upstream temperature sources (air conditioning exhaust, substations, etc.), the simulated data on airflow, temperature gradients, and pollutant diffusion paths within the factory area are transferred to other simulation models as boundary input conditions.

[0089] Air conditioning system simulation model: Based on the Modelica language, a modular simulation model is built, which includes air conditioning units, control systems, pipeline networks, etc. The simulation model of the factory building complex and the simulation model of the factory production environment provide input conditions for this model, and output air conditioning system operating parameters (such as chilled water flow, air volume, valve opening) and control signals (such as PID output value, air valve adjustment command).

[0090] Factory production environment simulation model: Simulates airflow organization, temperature, humidity, pressure, pollutant concentration, etc. in the production area. The simulation results of the factory building complex simulation model and the air conditioning system simulation model provide input boundary conditions for this model. It can obtain real-time values ​​of temperature, humidity, pressure difference, and particle concentration at each monitoring point in the clean room, as well as early warning data of risk points (such as local temperature exceeding the standard, insufficient airflow velocity).

[0091] The above three models achieve coupled simulation: the factory building complex simulation model is the top-level model, providing dynamic data of the external environment as input conditions for other model simulations; the air conditioning system simulation model and the factory production environment simulation model receive the output data of the factory building complex simulation model as input conditions, and at the same time, the two exchange the generated results as input to each other, forming a closed loop.

[0092] 6. Use the historical simulation data and historical operation and maintenance data corresponding to the nearby reference working condition point as the initial values ​​for the joint simulation platform to conduct simulation, and obtain the risk point data and simulation operation and maintenance data under the current weather variable factors.

[0093] Since climate fluctuations are not very drastic, historical simulation data can be used as initial values ​​to accelerate simulation convergence. The indoor environmental parameters and monitoring points calculated by the simulation are used as risk point data (such as whether there is localized overcooling / heating, excessive particle concentration, etc.) to issue early warnings for the production environment.

[0094] 7. Weight the simulated operation and maintenance data and the operation and maintenance data predicted by the random forest regression model to calculate the recommended operation and maintenance data under the current weather variable factors.

[0095] 1. Calculate the predicted values ​​y of the random forest regression model for weather variables at historical time stamps. {true} and the true value y {reg} Historical errors {reg} =y {true} -y {reg} If we assume that historical errors follow a certain probability distribution, then we further estimate the mean and variance of that probability distribution.

[0096] If we assume that the historical error of the random forest regression model follows a normal distribution, we can estimate its mean and variance to obtain the normal distribution N(μ). {reg} ,σ {reg} ), where μ {reg} The historical error mean, σ {reg} This represents the historical error variance.

[0097] 2. Calculate historical errors simulated on the co-simulation platform. {CAE} This refers to the historical error between the operation and maintenance data simulated by the co-simulation platform under historical operating conditions and the actual operation and maintenance data. It is assumed that this error follows a certain probability distribution, and its mean and variance are estimated. If actual historical error data is unavailable, a mean and variance are initially determined based on experience or preliminary experiments, and then adjusted based on more data to obtain a normal distribution N(μ) for the historical error of the co-simulation platform. {CAE} ,σ {CAE} ), where μ CAE The historical error mean, σ CAEThis represents the historical error variance.

[0098] 3. Calculate the weighting coefficients of the simulation data and predicted data based on the output results of the random forest regression model and the co-simulation platform: First, calculate the probability density of the predicted operation and maintenance data error based on the probability distribution (pdf). {reg} Probability density of simulation operation and maintenance data errors (pdf) {CAE} Sum of: pdf {total} =pdf {reg} +pdf {CAE} .

[0099] The weighting coefficients for the predicted operation and maintenance data corresponding to the random forest regression model are:

[0100]

[0101] The weighting coefficients for the simulation operation and maintenance data corresponding to the co-simulation platform are:

[0102]

[0103] 4. Prediction of the simulation operation and maintenance data from step six. {cfd} And random forest regression model to predict operational data {reg} Weighted calculations are performed to obtain the recommended operation and maintenance data (prediction) under the current weather variable factors. {final} :

[0104] prediction {final} =weight {reg} ×prediction {reg} +weight {cfd} ×prediction {cfd} .

[0105] 8. Upload the monitoring risk point data and recommended operation and maintenance data to the terminal to notify the operation and maintenance management personnel. The terminal includes, but is not limited to, the central control room, the operation and maintenance personnel's mobile phones, etc. The operation and maintenance personnel adjust the air conditioning system operation status according to the recommended operation and maintenance data or realize the automatic adjustment of the air conditioning system by pre-implanting corresponding functions in the automatic control system.

[0106] Based on the actual operating status of the air conditioning system, record the current risk point data, the operating point of the air conditioning system, and the actual operation and maintenance data, and upload them, along with the operation and maintenance data predicted by the random forest regression model and the historical simulation data of the joint simulation platform, to the operation and maintenance management database.

Claims

1. A method for intelligent control of air conditioning systems in industrial cleanrooms based on the prediction of production environment parameters, characterized in that, Includes the following steps: S1: Build an operation and maintenance management database, including multiple sets of historical data. Each set of historical data includes weather variable factors with historical timestamps, as well as historical predicted operation and maintenance data, historical simulation data, and historical operation and maintenance data corresponding to the weather variable factors. S2: Establish a random forest regression model between weather variables and historical operation and maintenance data; S3: Obtain current weather variables, use a random forest regression model to make predictions, obtain predicted operation and maintenance data, and store it in the operation and maintenance management database; S4: Based on the current weather variable factors, perform sensitivity analysis on the weather variable factors in the operation and maintenance management database, and use the operating point with the most similar sensitive conditions as the reference operating point near the current weather variable factor; S5: Build a joint simulation platform, including simulation models of the factory building complex, air conditioning system, and factory production environment; S6: Use the historical simulation data and historical operation and maintenance data corresponding to the nearby reference working condition point as the initial values ​​of the joint simulation platform to perform simulation, obtain the risk point data and simulation operation and maintenance data under the current weather variable factors, and store them in the operation and maintenance management database. S7: Weight the simulated operation and maintenance data and the predicted operation and maintenance data to obtain the recommended operation and maintenance data under the current weather variable factors, and store it in the operation and maintenance management database; S8: Upload risk point data and recommended operation and maintenance data to the terminal to enable automatic or manual adjustment of the air conditioning system; The factory building complex simulation model described in S5 is the top-level model, providing dynamic data of the external environment as the simulation input conditions for the air conditioning system simulation model and the factory production environment simulation model. The air conditioning system simulation model and the factory production environment simulation model receive the output data of the factory building complex simulation model as input conditions. The air conditioning system simulation model and the factory production environment simulation model exchange simulation results as input conditions for each other, forming a closed-loop joint simulation platform. S7 describes weighting the simulated operation and maintenance data and the predicted operation and maintenance data to obtain recommended operation and maintenance data under the current weather variable factors, including: Calculate the weighting coefficients for simulated operation and maintenance data and predicted operation and maintenance data. Then, the simulation operation and maintenance data described in S6 will be used. The predictive maintenance data of the random forest regression model described in S3 Weighted calculations are performed to obtain recommended operation and maintenance data under the current weather variable factors. : ； The weighting coefficients of the simulated operation and maintenance data and the predicted operation and maintenance data They are respectively: ， ， in, To predict the probability density of operational data errors, To simulate the probability density of operational and maintenance data errors, .

2. The intelligent control method for an industrial cleanroom air conditioning system according to claim 1, characterized in that, The weather variables mentioned in S1 include air pressure, weather conditions, precipitation probability, precipitation amount, temperature, perceived temperature, relative humidity, wind direction, wind speed, gusts, and cloud cover; the operating point refers to the operation and maintenance data of the air conditioning system under specific weather variable conditions.

3. The intelligent control method for an industrial cleanroom air conditioning system according to claim 2, characterized in that, The historical simulation data mentioned in S1 includes historical risk point data and historical simulation operation and maintenance data; the operation and maintenance data of the air conditioning system includes cooling water temperature, cooling water flow rate, chilled water temperature, chilled water flow rate, supply air temperature, supply air volume, return air volume, fresh air volume, cleanroom temperature, cleanroom humidity, cleanroom pressure and particle count, water valve opening, air valve opening, equipment start-up and shutdown records, fault records, alarm information and abnormal records.

4. The intelligent control method for an industrial cleanroom air conditioning system according to claim 1, characterized in that, S2 describes establishing a random forest regression model between weather variables and historical operation and maintenance data, including: S201. Data preprocessing: Quantify, clean, and standardize the weather variable factors and historical operation and maintenance data coding methods in each set of historical data. S202. Establish a random forest regression model: Use the preprocessed weather variables in each set of data as feature variables X, and the historical operation and maintenance data as target variables Y, and train the random forest regression model to obtain the trained random forest regression model.

5. The intelligent control method for an industrial cleanroom air conditioning system according to claim 1, characterized in that, S4 obtains the reference operating point through sensitivity analysis as follows: S401. Use the trained random forest model to calculate the feature importance score of weather variable factors in each group of historical data, and calculate the weight of weather variable factors based on feature importance. S402. Match the current weather variable factors with the weather variable factors in each set of historical data. For each historical data point, use weights to calculate the weighted Euclidean distance between the current weather variable factors and the historical weather variable factors, and calculate the weighted Euclidean distance between the current expected operating point and the historical operating point. S403. Traverse the historical data, calculate the comprehensive distance of each set of historical data, and select the working point corresponding to the historical data with the smallest comprehensive distance as the reference working point for the current weather variable factor.

6. The intelligent control method for an industrial cleanroom air conditioning system according to claim 1, characterized in that, The method for obtaining the probability density of predicted operation and maintenance data error and the probability density of simulated operation and maintenance data error is as follows: calculate the historical error between historical predicted operation and maintenance data and historical operation and maintenance data corresponding to each historical weather variable factor in the operation and maintenance management database, and set the predicted operation and maintenance data error to follow a probability distribution; calculate the historical error between historical simulated data and historical operation and maintenance data corresponding to each historical weather variable factor in the operation and maintenance management database, and set the simulated operation and maintenance data error to follow a probability distribution. Calculate the probability density of predicted operational data errors based on probability distribution. Probability density of simulation operation and maintenance data errors .

7. An intelligent control system for an industrial cleanroom air conditioning system based on production environment parameter prediction, the system being used to execute the intelligent control method for an industrial cleanroom air conditioning system based on production environment parameter prediction as described in claim 1, characterized in that, include: Database module: Build an operation and maintenance management database, including multiple sets of historical data. Each set of historical data includes weather variable factors with historical timestamps, as well as historical predicted operation and maintenance data, historical simulation data, and historical operation and maintenance data corresponding to the weather variable factors. Predictive model building module: Establishes a random forest regression model between weather variables and historical operation and maintenance data; Predictive Operation and Maintenance Data Generation Module: Obtains current weather variables and uses a random forest regression model to make predictions, thereby obtaining predictive operation and maintenance data; Reference operating condition point acquisition module: Based on the current weather variable factors, perform sensitivity analysis on the weather variable factors in the operation and maintenance management database, and use the operating condition point with the most similar sensitive conditions as the reference operating condition point near the current weather variable factor; Co-simulation platform construction module: Builds a co-simulation platform, including simulation models of the factory building complex, air conditioning system, and factory production environment; Simulation data generation module: It uses historical simulation data and historical operation and maintenance data corresponding to nearby reference working conditions as the initial values ​​of the joint simulation platform to simulate and obtain risk point data and simulation operation and maintenance data under the current weather variable factors, and stores them in the operation and maintenance management database. Recommended Operation and Maintenance Data Generation Module: This module performs weighted calculations on simulated and predicted operation and maintenance data to obtain recommended operation and maintenance data under the current weather variable factors. Data upload module: Uploads risk point data and recommended operation and maintenance data to the terminal to enable automatic or manual adjustment of the air conditioning system.

Images