Intelligent clean air conditioner automatic control system and integrated control method

Abstract

The invention discloses an intelligent clean air conditioner automatic control system and an integrated control method, and belongs to the technical field of clean environment control. In order to solve the problems of dispersed control, high energy consumption and insufficient dynamic response of an existing clean air conditioning system, an intelligent control algorithm and hardware facilities are integrated, a multi-parameter collaborative optimization model is constructed, and the temperature, humidity, airflow velocity and pressure of a clean workshop are accurately regulated and controlled in real time. The system comprises a central controller, a sensor network, an actuator and a control software module, an air valve and variable air volume equipment are dynamically adjusted through data fusion and a machine learning algorithm, wind resistance and turbulence are reduced, and airflow organization efficiency is optimized. Meanwhile, the system supports remote monitoring, fault self-diagnosis and self-adaptive energy-saving modes, and the operation stability and the energy efficiency ratio are remarkably improved. According to the integrated control method, the problems that a traditional system is insufficient in intelligence and high in maintenance cost are solved, and remarkable energy-saving benefits and environmental adaptability are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of clean environment control, in particular to an intelligent clean air conditioning automatic control system and an integrated control method. Background Art

[0002] In the field of clean environment control (such as medicine, electronic manufacturing, biological laboratories, etc.), the air conditioning system needs to maintain constant temperature, humidity, pressure difference and air flow speed to ensure the cleanliness and stability of the production environment. Traditional clean air conditioning control systems mostly use classic PID control algorithms. Although they have basic adjustment capabilities, they face the following technical bottlenecks under complex dynamic conditions:

[0003] 1. The balance between control accuracy and energy consumption

[0004] Traditional PID parameters are fixed and cannot be adjusted in real time according to environmental fluctuations, resulting in a sharp increase in energy consumption during high-precision control (such as frequent start and stop of fans), or sacrificing control stability for energy saving (such as accumulation of temperature and humidity deviations).

[0005] 2. Lack of multi-parameter collaborative optimization

[0006] Existing technologies mostly focus on independent control of a single parameter (such as temperature), lacking the integrated analysis and coordinated regulation of multiple variables such as temperature, humidity, pressure difference, wind speed, etc., resulting in low airflow organization efficiency (η<0.8).

[0007] 3. Insufficient fault diagnosis and emergency response capabilities

[0008] Traditional systems rely on threshold alarms and cannot predict potential failures (such as sensor drift and actuator sticking). In addition, the delay in switching to safe mode under emergency conditions such as a sudden drop in pressure differential is high (>1 second), which may cause the clean environment to fail.

[0009] 4. Limitations of intelligence and adaptive capabilities

[0010] Existing solutions rarely combine machine learning with multi-objective optimization algorithms, making it difficult to dynamically optimize PID parameters and energy consumption weights, resulting in a year-on-year decline in the long-term operating energy efficiency ratio (COP).

[0011] As the goal is advanced, clean air conditioning systems need to simultaneously meet the needs of high-precision control, low-energy operation, and intelligent operation and maintenance. Although some studies have attempted to introduce fuzzy control or data fusion technology (such as the Chinese patent "A multi-parameter coordinated control method for air conditioning"), the following defects still exist:

[0012] The fuzzy rule base is fixed and cannot be updated online to adapt to changes in working conditions;

[0013] The optimization goal is single, and multi-dimensional indicators such as energy consumption, control accuracy and equipment life are not coordinated;

[0014] The edge computing and cloud collaboration capabilities are weak, and the utilization rate of historical data is low.

[0015] In view of the above problems, the present invention proposes an intelligent clean air-conditioning automatic control system and an integrated control method, which improve the control performance and intelligent level of the clean air-conditioning system through dynamic data fusion, fuzzy PID parameter self-adaptive adjustment, multi-objective optimization model, and edge-cloud collaborative management. Summary of the Invention

[0016] In view of the existing problems above, the present invention is proposed.

[0017] Therefore, the intelligent clean air-conditioning automatic control system and the integrated control method provided by the present invention solve the problems of the balance between control accuracy and energy consumption, the lack of multi-parameter collaborative optimization, and the insufficient fault diagnosis and emergency response capabilities.

[0018] To solve the above technical problems, the present invention provides the following technical solutions:

[0019] In a first aspect, the present invention provides an intelligent clean air-conditioning automatic control system, including:

[0020] A central controller, a distributed sensor network, an actuator, and a control software module;

[0021] The central controller adopts a multi-core processor architecture, has a built-in real-time operating system (RTOS), and is configured with communication interfaces including RS-485, CAN bus, and Ethernet;

[0022] The distributed sensor network is composed of a temperature sensor (accuracy ±0.1°C), a humidity sensor (accuracy ±1%RH), a differential pressure sensor (range 0-500Pa), and an air velocity sensor (range 0-10m / s), which are installed at the four corners and the core area of the clean workshop, with a sampling frequency of 10Hz to collect the environmental parameters of the clean workshop in real time;

[0023] The actuator includes an electric proportional air valve (opening resolution 0.1%), a variable-frequency fan (speed regulation range 10%-100%), and a PID-regulated humidifier / dehumidifier (control accuracy ±2%RH), and its opening or speed is adjusted by the control signal output by the central controller;

[0024] The control software module is deployed on the central controller, integrates data fusion, fuzzy PID control, and multi-objective optimization algorithms, and supports remote interaction through the Modbus-TCP protocol;

[0025] The control software module performs weighted processing on multi-sensor data through a data fusion algorithm, and the weight distribution formula is:

[0026]

[0027] Among them, σ i is the measurement standard deviation of the i-th type of sensor.

[0028] As a preferred solution of the intelligent clean air-conditioning automatic control system described in the present invention, wherein: the central controller adopts a fuzzy PID control algorithm, and its output control quantity u(t) is:

[0029]

[0030] Among them, e(t) = P target -P real is the environmental parameter deviation, P target is the target environmental parameter of the clean workshop, P real is the actual environmental parameter of the clean workshop, K p , K i , K d is the gain coefficient adjusted dynamically, K p is the proportional gain, K i is the integral gain), K d is the differential gain, K p (t) ∈ [0.5, 2.0], K i (t) ∈ [0.01, 0.1], K d (t) ∈ [0.05, 0.5].

[0031] As a preferred solution of the intelligent clean air-conditioning automatic control system described in the present invention, wherein: the objective function of the multi-objective optimization model of the control software module is:

[0032]

[0033] The constraint conditions include: V ∈ [0.2, 0.8] m / s, T ∈ [20, 26];

[0034] In the formula, E base is the reference energy consumption (unit: kWh), E current is the current cycle energy consumption (unit: kWh), P k,range is the reference environmental parameter of the clean workshop, α = 0.6, β = 0.4, and the weights are optimized every 30 minutes through the NSGA-II algorithm.

[0035] As a preferred solution of the intelligent clean air-conditioning automatic control system described in the present invention, wherein: the relationship between the air volume Q and the opening θ of the electric proportional air valve is:

[0036]

[0037] Among them,

[0038] ρ = 1.2 kg / m, the opening θ ∈ [0%, 100%], and the accuracy is guaranteed by a 12-bit DAC module.

[0039] As a preferred solution of the intelligent clean air-conditioning integrated control method described in the present invention, it includes the following steps:

[0040] S1. The environmental parameters P are collected in real time at a frequency of 10 Hz through a distributed sensor network, and the noise is eliminated through Kalman filtering.

[0041] S2. Calculate the weighted comprehensive value of environmental parameters based on the data fusion algorithm The weight W i Is dynamically updated according to the health status of the sensor.

[0042] S3. Query the fuzzy rule table according to e(t) and γ(t), and output the real-time control parameters K p , K i , K d , and use the fuzzy PID controller to generate the control signal U = {u 1 , u 2 , u 3}, corresponding to the air valve opening, the fan speed, and the humidification power respectively, and dynamically adjust the actuator according to the control signal U.

[0043] S4. Call the NSGA-II algorithm every 30 minutes to optimize α and β, update the PID parameter threshold, and update the control parameter K through the multi-objective optimization model p , K i , K d , and feedback it to step S3;

[0044] S5. Steps S1 - S4 are executed in sequence in a loop, and the control period is 100 ms; when an emergency condition (such as a sudden drop in pressure difference exceeding 20%) is detected, the interruption mechanism is triggered, and S4 is skipped and the preset safety parameters are directly adopted; the historical data is stored in the edge computing node, supporting retrospective analysis and model retraining, and the data retention period is 30 days.

[0045] As a preferred solution of the intelligent clean air-conditioning integrated control method described in the present invention, the step of collecting the environmental parameters P in real time at a frequency of 10 Hz through a distributed sensor network and eliminating the noise through Kalman filtering includes the following steps:

[0046] Synchronously collect the temperature T, humidity H, pressure difference ΔP, and wind speed V in the clean workshop at a frequency of 10 Hz through a distributed sensor network. Among them, the temperature sensor uses a PT100 platinum resistance, the humidity sensor uses a capacitive polymer film, the pressure difference sensor is a micro-electro-mechanical system (MEMS) structure, and the wind speed sensor is a hot wire type.

[0047] Perform Kalman filter preprocessing on the original data, and the filtering formula is:

[0048]

[0049] In the formula, is the filtering environment parameter at time k, and P k is the environment parameter at time k, is the filtering environment parameter at time k - 1, and K k is the Kalman gain, H is the observation matrix, the noise covariance matrix Q = 0.01I, R = 0.05I, where I is the identity matrix;

[0050] The data storage format is 32-bit floating-point number, and CRC-16 check is used to ensure transmission integrity.

[0051] As a preferred solution of the intelligent clean air conditioner integrated control method described in the present invention, wherein: calculating the weighted environmental parameter comprehensive value based on the data fusion algorithm The weight W i is dynamically updated according to the sensor health status, including the following steps,

[0052] Dynamically adjust the weight W according to the sensor health status i , and the health status score S i The calculation formula is

[0053]

[0054] where R is the number of recent failures;

[0055] The corrected fusion weight is W i ′ = W i ·S i , and the comprehensive value is calculated as:

[0056]

[0057] If a certain sensor S i < 0.5, then trigger the redundant sensor switch and update the fusion calculation logic.

[0058] As a preferred solution of the intelligent clean air conditioner integrated control method described in the present invention, wherein: querying the fuzzy rule table according to e(t) and γ(t) and outputting the real-time control parameters K p , K i , K d , and using the fuzzy PID controller to generate the control signal U = {u 1 , u 2 , u 3}, corresponding to the damper opening, the fan speed, and the humidification power respectively, dynamically adjust the actuator according to the control signal U, including the following steps:

[0059] Fuzzify the input variables: the environmental parameter deviation e(t) and the deviation change rate γ(t), which are divided into 7 fuzzy sets (NB, NM, NS, ZO, PS, PM, PB);

[0060] The fuzzy rule table adopts the Mamdani inference mechanism, and an example rule is:

[0061] IF e(t) is PB AND γ(t) is NB THEN K p is PB, K i is NB, K d is PS;

[0062] Defuzzify using the centroid method to output the precise PID parameters:

[0063]

[0064] where μ j is the activation degree of the j-th fuzzy rule;

[0065] The calculation period of the control signal U is 100 ms, and it is output to the actuator through the DA conversion module.

[0066] As a preferred solution of the intelligent clean air-conditioning integrated control method described in the present invention, wherein: the NSGA-II algorithm is called every 30 minutes to optimize α, β, and update the PID parameter threshold, and the control parameter K is updated through the multi-objective optimization model p , K i , K d , and feedback to step S3, including the following steps:

[0067] Use the NSGA-II algorithm to optimize the objective function weights α, β, the population size is 50, the crossover probability is 0.8, and the mutation probability is 0.1;

[0068] Generate the Pareto front solution set every 30 minutes, select the comprehensive optimal solution of energy consumption and parameter deviation, and update the PID parameter threshold:

[0069]

[0070] where, is the control parameter threshold before update, is the control parameter threshold after update, δ is the adaptive learning rate coefficient, which is used to control the update amplitude of the PID parameters, and ΔE is the energy consumption optimization rate, indicating the current cycle energy consumption E current and the reference energy consumption E baseRelative difference;

[0071] The optimization results are uploaded to the cloud monitoring platform through the OPC UA protocol and a visualization report is generated.

[0072] As a preferred solution of the intelligent clean air conditioner integrated control method described in the present invention, wherein: the steps S1-S4 are executed in sequence and cyclically, and the control period is 100 ms; when an emergency condition (such as a sudden drop in pressure difference exceeding 20%) is detected, an interruption mechanism is triggered, and S4 is skipped and preset safety parameters are directly adopted; historical data is stored in the edge computing node, supporting retrospective analysis and model retraining, and the data retention period is 30 days, including the following steps.

[0073] (a) Periodic control loop execution:

[0074] Steps S1 to S4 run in sequence and cyclically, the control period is fixed at 100 ms, and is triggered by a high-precision hardware timer (model STM32F4xx, clock source error ±0.001%) built into the central controller;

[0075] Within each period, the data acquisition time ≤ 5 ms, the data fusion and fuzzy PID calculation time ≤ 80 ms, and the remaining time is used for asynchronous iterative calculation of multi-objective optimization;

[0076] (b) Emergency condition interruption detection and response:

[0077] The pressure difference parameter ΔP is monitored in real time. If ∣ΔP t -ΔP t-1 ∣ / ΔP t-1 ≥ 20% lasts for 3 sampling periods (30 ms), then it is determined as an emergency condition of sudden drop in pressure difference;

[0078] Trigger the interrupt service routine (ISR), immediately pause the current control period, skip the optimization step S4, and call the preset safety parameter set U safe ={θ=60%, N=50%, P humid =30%};

[0079] In the safety mode, the actuator (103) receives U safe at a fixed frequency of 10 Hz until the environmental parameters return to stability (ΔP volatility < 5% for 10 seconds);

[0080] (c) Edge data storage and intelligent management:

[0081] Historical data is stored in the edge computing node, configured with an ARM Cortex-A72 processor + NVMe SSD. After being encrypted by timestamp (AES-256), it is written into the time series database (InfluxDB). The stored fields include T, H, ΔP, V, U, η and the fault flag bit. Among them, T is the real-time temperature data in the clean workshop, with the unit of degree Celsius (°C), collected by the temperature sensor; H is the real-time relative humidity data in the clean workshop, with the unit of percentage (%), collected by the humidity sensor; ΔP is the pressure difference data between the clean workshop and the external environment, with the unit of Pascal (Pa), measured by the pressure difference sensor, used to monitor the air flow direction and airtightness; V is the air flow velocity data in the clean workshop, with the unit of meter per second (m / s), collected by the anemometer, used to evaluate the air flow organization efficiency; U is the set of control signals of the actuator, output by the central controller, used to dynamically adjust the system operation state; η is the air flow uniformity index;

[0082] The data retention policy is a rolling deletion mechanism, which automatically clears the records 30 days ago and generates a compressed archive file (ZIP + CRC32 check) and uploads it to the cloud;

[0083] It supports the retraining of the model based on the TensorFlow Lite framework. Every 7 days, 10% of the samples (uniform distribution) are extracted from the historical data to update the weights of the LSTM prediction model. The optimized model is updated to the central controller via OTA.

[0084] In a second aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, it implements any step of an intelligent clean air conditioner integrated control method as described in the first aspect of the present invention.

[0085] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, it implements any step of an intelligent clean air conditioner integrated control method as described in the first aspect of the present invention.

[0086] The beneficial effects of the present invention are:

[0087] Based on Kalman filtering and dynamic data fusion, the present invention effectively eliminates sensor noise and reduces the comprehensive parameter measurement error to within ±1%. Through adaptive fuzzy PID control, the control accuracy and dynamic response ability are comprehensively improved. The NSGA-II algorithm is used to dynamically balance the energy consumption and accuracy objectives, enabling the system's coefficient of performance (COP) to be increased by 15%-20% and the annual comprehensive energy consumption to be reduced by approximately 12%-18%. The adaptive energy-saving mode is adopted. When the environmental parameters are stable, it automatically switches to the low-power state (the fan speed drops to 40%), and the energy-saving efficiency is increased by 25%. Based on the LSTM prediction model and the 3σ criterion, the fault detection accuracy rate is ≥95%, and the interruption response delay under emergency conditions (such as a sudden 20% drop in pressure difference) is ≤30 ms, avoiding the misjudgment and delay (>1 second) of traditional threshold alarms. Historical data is encrypted and stored (AES-256) and supports model retraining (updated once every 7 days), reducing the operation and maintenance cost by 30% and extending the equipment life by 10%-15%. By predicting the air flow uniformity index through a BP neural network (η≥0.9), compared with traditional empirical regulation (η≈0.7), the coverage rate of the clean area is increased by 20%-30%. Through the linkage adjustment of parameters such as temperature and humidity, pressure difference, and wind speed, the compliance rate of the clean workshop with the ISO 14644 standard is increased from 80% to over 98%. Through the Modbus-TCP / OPC UA protocol, remote real-time monitoring is achieved, and the firmware and algorithm can be upgraded online, reducing the downtime for maintenance by more than 50%. Through dynamic data fusion, fuzzy PID adaptive control, multi-objective optimization, and intelligent operation and maintenance mechanisms, the present invention systematically solves the defects of traditional clean air-conditioning systems in terms of control accuracy, energy consumption, reliability, and intelligent level. Verified by experiments, it can significantly reduce the operation and maintenance cost, improve the coefficient of performance, and provide a stable and reliable environmental guarantee for high-cleanliness industrial scenarios, with significant economic and social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0089] Figure 1 It is a schematic diagram of an intelligent clean air-conditioning automatic control system in Embodiment 1;

[0090] Figure 2 It is a flowchart of an intelligent clean air-conditioning integrated control method in Embodiment 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0091] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0092] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0093] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0094] Embodiment 1, referring to Figure 1 , is the first embodiment of the present invention. This embodiment provides an intelligent clean air-conditioning automatic control system, including:

[0095] A central controller 101, a distributed sensor network 102, an actuator 103, and a control software module 104;

[0096] The central controller 101 adopts a multi-core processor architecture, with a built-in real-time operating system (RTOS), and is configured with communication interfaces including RS-485, CAN bus, and Ethernet;

[0097] The distributed sensor network 102 consists of a temperature sensor (accuracy ±0.1°C), a humidity sensor (accuracy ±1%RH), a differential pressure sensor (range 0 - 500 Pa), and an air velocity sensor (range 0 - 10 m / s). It is installed at the four corners and the core area of the clean workshop, with a sampling frequency of 10 Hz, and real-time collects the environmental parameters of the clean workshop;

[0098] The actuator 103 includes an electric proportional air valve A1 with an opening resolution of 0.1%, a variable-frequency fan A2 with a speed regulation range of 10% - 100%, and a PID-regulated humidifier / dehumidifier A3 with a control accuracy of ±2%RH. Its opening or speed is adjusted by the control signal output by the central controller 101;

[0099] The control software module 104 is deployed on the central controller 101, integrates data fusion, fuzzy PID control, and multi-objective optimization algorithms, and supports remote interaction via the Modbus-TCP protocol;

[0100] The control software module 104 performs weighted processing on multi-sensor data through a data fusion algorithm. The weight distribution formula is:

[0101]

[0102] where σ iis the measurement standard deviation of the i-th type of sensor.

[0103] The central controller 101 adopts a fuzzy PID control algorithm, and its output control quantity u(t) is:

[0104]

[0105] where e(t) = P target - P real is the environmental parameter deviation, P target is the target environmental parameter of the clean workshop, P real is the actual environmental parameter of the clean workshop, K p , K i , K d is the gain coefficient adjusted dynamically, K p is the proportional gain, K i is the integral gain), K d is the differential gain, K p (t) ∈ [0.5, 2.0], K i (t) ∈ [0.01, 0.1], K d (t) ∈ [0.05, 0.5].

[0106] The objective function of the multi-objective optimization model of the control software module 104 is:

[0107]

[0108] The constraint conditions include: V ∈ [0.2, 0.8] m / s, T ∈ [20, 26];

[0109] In the formula, E base is the reference energy consumption (unit: kWh), E current is the current cycle energy consumption (unit: kWh), P k,range is the reference environmental parameter of the clean workshop, α = 0.6, β = 0.4, and the weights are optimized every 30 minutes by the NSGA-II algorithm.

[0110] The relationship between the air volume Q and the opening θ of the electric proportional air valve A1 is:

[0111]

[0112] where

[0113] ρ = 1.2 kg / m, the opening θ ∈ [0%, 100%], and the accuracy is guaranteed by a 12-bit DAC module.

[0114] Example 2, referring to Figure 2 , is the second embodiment of the present invention, and this embodiment provides an intelligent

[0115] Clean

[0116] Integrated control method for clean air conditioners, including the following steps,

[0117] S1. The environmental parameters P are collected in real time at a frequency of 10 Hz through the distributed sensor network 102, and the noise is eliminated through Kalman filtering;

[0118] Data acquisition process:

[0119] The temperature T, humidity H, differential pressure ΔP and wind speed V in the clean workshop are synchronously collected through the distributed sensor network 102 at a frequency of 10 Hz. Among them, the temperature sensor uses a PT100 platinum resistance, the humidity sensor uses a capacitive polymer film, the differential pressure sensor is a microelectromechanical system (MEMS) structure, and the wind speed sensor is a hot wire type;

[0120] Sensor selection:

[0121] Temperature sensor: Use a PT100 platinum resistance (model: Omega PT100-1K), with an accuracy of ±0.1 °C and a range of -20 °C to 80 °C (model: Omega PT100-1K). It is installed at the four corners of the clean workshop and around the core equipment (8 nodes are arranged per 100 ㎡), and is connected to the central controller through a shielded cable;

[0122] Humidity sensor: Select a Honeywell HIH-4000 capacitive sensor, with an accuracy of ±1%, a range of 0-100%, installed in the air return opening of the air conditioner and inside the air supply duct (2 nodes per duct), and avoid direct contact with condensate;

[0123] Differential pressure sensor: Use a Sensirion SDP800 MEMS sensor, with an accuracy of ±1 Pa and a range of 0-500 Pa. It is installed on both sides of the partition wall between the clean room and the buffer room (1 pair per partition wall), and the data is transmitted through the RS-485 bus;

[0124] Wind speed sensor: TSI 8475 hot wire anemometer, with a resolution of 0.01 m / s and a range of 0-10 m / s. It is installed 0.5 meters below the high-efficiency air supply opening (1 per air outlet), and the sampling frequency is 50 Hz;

[0125] Synchronization mechanism:

[0126] The distributed sensor network synchronously collects data at a frequency of 10 Hz through the LoRa wireless communication protocol

[0127] (Chinese

[0128] The heart frequency is 433 MHz and the bandwidth is 125 kHz, and it is transmitted to the central controller (STM32F407VG); the central controller triggers 10 Hz synchronous acquisition through the hardware timer (TIM2, 72 MHz clock), and each sensor node uses the GPS module (ublox NEO-M8N) to achieve microsecond-level time synchronization.

[0129] Preprocessing method: Kalman filtering is used to eliminate noise, and the filtering parameters are set as the process noise covariance Q = 0.01I and the observation noise covariance R = 0.05I. The output is the smoothed environmental parameter set P = {T, H, ΔP, V}, where T is the real-time temperature in the clean workshop, in degrees Celsius (°C), H is the real-time relative humidity in the clean workshop, in percentage (%), ΔP is the pressure difference between the clean workshop and the adjacent area, in Pascals (Pa), and V is the air flow velocity in the clean workshop, in meters per second (m / s).

[0130] Implementation of Kalman filtering:

[0131] The original data is preprocessed by Kalman filtering, and the filtering formula is:

[0132]

[0133] In the formula, is the filtered environmental parameter at time k, P k is the environmental parameter at time k, is the filtered environmental parameter at time k-1, K k is the Kalman gain, H is the observation matrix, the noise covariance matrices Q = 0.01I and R = 0.05I, where I is the identity matrix;

[0134] Data integrity check:

[0135] The data storage format is 32-bit floating-point number, and CRC-16 check is used to ensure the transmission integrity. If the check fails, the retransmission mechanism is triggered (up to 3 times) to ensure that the transmission packet loss rate ≤ 0.5%.

[0136] S2. Calculate the weighted comprehensive value of environmental parameters based on the data fusion algorithm Weight W i is dynamically updated according to the health status of the sensor;

[0137] Dynamic weight allocation: The initial weight is

[0138] where σ T = 0.05, σ H = 0.5, σ ΔP = 2, σ V = 0.1;

[0139] According to the sensor health status score where R is the number of recent failures; the corrected fusion weight is W i ′ = W i ·S i .

[0140] Redundancy switching mechanism:

[0141] If a certain sensor S i < 0.5, then trigger the redundancy sensor switch, automatically enable the standby sensor (such as the dual-redundancy sensor group installed at the four corners), and recalculate the fusion comprehensive value

[0142] Verification and calibration:

[0143] Offline calibration: Use a standard instrument (Fluke 754 process calibrator) to calibrate the sensor on-site every week, and correct the sensors with a deviation exceeding ±2%.

[0144] Online verification: Compare the fusion value P fused with the manual measurement value every 1 hour. If the deviation > 3%, then trigger an alarm and record the log.

[0145] S3. Query the fuzzy rule table according to e(t) and γ(t), and output the real-time control parameter K p , K i , K d , and use the fuzzy PID controller to generate the control signal U = {u 1 , u 2 , u 3}, corresponding to the damper opening, fan speed, and humidification power respectively, and dynamically adjust the actuator according to the control signal U;

[0146] Fuzzify the input variables: The environmental parameter deviation e(t) = P target -P real and the deviation change rate γ(t), which are divided into 7 fuzzy sets (NB, NM, NS, ZO, PS, PM, PB);

[0147] Output variable: PID gain coefficient K p , K i , K d , and the ranges are K p ∈[0.5, 2.0], K i ∈[0.01, 0.1], K d ∈[0.05, 0.5];

[0148] The fuzzy rule table adopts the Mamdani inference mechanism, and an example rule is:

[0149] If e(t) is PB and γ(t) is NB then K p is PB, K i is NB, K d is PS;

[0150] The defuzzification adopts the centroid method to output the accurate PID parameters:

[0151]

[0152] where μ j is the activation degree of the j-th fuzzy rule, determined by the minimum value of the input membership degree;

[0153] The calculation period of the control signal U is 100 ms. The control signal U drives the electric air valve (opening resolution 0.1%), the variable-frequency fan (speed regulation range 10% - 100%), and the humidifier (control accuracy ±2% RH) respectively, and is output to the actuator (103) through the DA conversion module.

[0154] The control signal U = {u 1 , u 2 , u 3} is generated:

[0155] Opening of the electric air valve: Output resolution 0.1%.

[0156] Speed of the variable-frequency fan: where D = 0.5 m, η = 0.75, speed regulation error < ±1%, Q target is the target air volume of the electric proportional air valve.

[0157] Humidification power: u 3 = sat(K p ·(H target - H real ), 0%, 100%), and the saturation function limits the output range.

[0158] S4. Call the NSGA-II algorithm to optimize α and β every 30 minutes, update the PID parameter thresholds, update the control parameters K p , K i , K d , and feedback to step S3;

[0159] Optimization process for updating control parameters by the multi-objective optimization model:

[0160] Objective function:

[0161]

[0162] The constraint conditions include: V ∈ [0.2, 0.8] m / s, T ∈ [20, 26];

[0163] In the formula, E base is the reference energy consumption (unit: kWh), E current is the current cycle energy consumption (unit: kWh), P k,range is the reference environmental parameter of the clean workshop, α = 0.6, β = 0.4, and the weights are optimized every 30 minutes by the NSGA-II algorithm.

[0164] Algorithm implementation:

[0165] The NSGA-II algorithm is used to optimize the weights α and β of the objective function, α ∈ [0, 1], β ∈ [0, 1], the population size is 50, the crossover probability is 0.8, and the mutation probability is 0.1;

[0166] Calculate the non-dominated sorting and crowding distance every 30 minutes to generate the Pareto front solution set, select the comprehensive optimal solution of energy consumption and parameter deviation, select the top 10% individuals as the optimal solution set, and update the PID parameter threshold:

[0167]

[0168] Among them, is the control parameter threshold before update, is the control parameter threshold after update, δ is the adaptive learning rate coefficient, used to control the amplitude of PID parameter update, ΔE is the energy consumption optimization rate, indicating the relative difference between the current cycle energy consumption E current and the reference energy consumption E base ;

[0169]

[0170] The optimization results are uploaded to the cloud monitoring platform through the OPC UA protocol and a visualization report is generated.

[0171] S5. Steps S1 - S4 are executed in sequence and loop, the control period is 100 ms; when an emergency condition is detected (such as a sudden drop in pressure difference exceeding 20%), the interruption mechanism is triggered, and S4 is skipped and the preset safety parameters are directly adopted; the historical data is stored in the edge computing node 105, supporting retrospective analysis and model retraining, and the data retention period is 30 days.

[0172] (a) Periodic control loop execution:

[0173] Steps S1 to S4 are executed in sequence and loop, the control period is fixed at 100 ms, triggered by the high-precision hardware timer (model STM32F407VG, clock source error ±0.001%) built into the central controller 101, and the interrupt priority is configured as the highest level;

[0174] Within each cycle, the data acquisition time consumption ≤ 5 ms, the data fusion and fuzzy PID calculation time consumption ≤ 80 ms, and the remaining time is used for the asynchronous iterative calculation of multi-objective optimization;

[0175] (b) Emergency condition interruption detection and response:

[0176] Real-time monitor the differential pressure parameter ΔP. If ∣ΔP t -ΔP t-1 ∣ / ΔP t-1 ≥ 20% lasts for 3 sampling cycles (30 ms), then it is determined as an emergency condition of sudden drop in differential pressure;

[0177] Trigger the interrupt service routine (ISR), immediately pause the current control cycle, skip the optimization step S4, and call the preset safety parameter set U safe ={θ = 60%, N = 50%, P humid = 30%}; Stored by the EEPROM (model: AT24C256) to ensure that the parameters are not lost after power-off;

[0178] In the safety mode, the actuator 103 receives U at a fixed frequency of 10 Hz safe , until the environmental parameters return to stability (ΔP volatility < 5% lasts for 10 seconds);

[0179] (c) Edge data storage and intelligent management:

[0180] Edge storage configuration:

[0181] Hardware: NVIDIA Jetson Nano (4GB RAM) + 1TB NVMe SSD (model: Samsung 970EVO).

[0182] Software: The data is stored in the InfluxDB time series structure, and the encryption adopts the AES-256-CBC mode, and the key is rotated every 24 hours.

[0183] Historical data is stored in the edge computing node 105. After being encrypted by timestamp (AES-256), it is written into the time series database (InfluxDB). The stored fields include T, H, ΔP, V, U, η and the fault flag bit. Among them, T is the real-time temperature data in the clean workshop, with the unit of degree Celsius (°C), collected by the temperature sensor; H is the real-time relative humidity data in the clean workshop, with the unit of percentage (%), collected by the humidity sensor; ΔP is the differential pressure data between the clean workshop and the external environment, with the unit of Pascal (Pa), measured by the differential pressure sensor, used to monitor the air flow direction and airtightness; V is the air flow velocity data in the clean workshop, with the unit of meter per second (m / s), collected by the anemometer, used to evaluate the air flow organization efficiency; U is the set of control signals of the actuator, output by the central controller, used to dynamically adjust the system operation state; η is the air flow uniformity index;

[0184] The data retention policy is a rolling deletion mechanism, which automatically clears the records 30 days ago and generates a compressed archive file (ZIP + CRC32 checksum) and uploads it to the cloud;

[0185] Model retraining:

[0186] Every 7 days, 10% of the samples (about 300,000 records) are randomly selected from the historical data, and the LSTM prediction model (128 nodes in the hidden layer) is trained using the TensorFlow Lite framework. After the training error < 2%, it is pushed to the central controller through OTA.

[0187] This embodiment also provides a computer device, which is applicable to a situation of an intelligent integrated control method for a clean air conditioner, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement an intelligent integrated control method for a clean air conditioner as proposed in the above embodiment.

[0188] The computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0189] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the intelligent clean air conditioner integrated control method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc.

[0190] In summary, based on Kalman filtering and dynamic data fusion, the present invention effectively eliminates sensor noise and reduces the comprehensive parameter measurement error to within ±1%. Through adaptive fuzzy PID control, the control accuracy and dynamic response ability are comprehensively improved. The NSGA-II algorithm is used to dynamically balance the energy consumption and accuracy targets, increasing the coefficient of performance (COP) of the system by 15%-20% and reducing the annual comprehensive energy consumption by about 12%-18%. The adaptive energy-saving mode is adopted. When the environmental parameters are stable, it automatically switches to the low-power state (the fan speed drops to 40%), and the energy-saving efficiency is increased by 25%. Based on the LSTM prediction model and the 3σ criterion, the fault detection accuracy rate is ≥95%, and the interruption response delay under emergency conditions (such as a sudden 20% drop in differential pressure) is ≤30 ms, avoiding the misjudgment and delay (>1 second) of traditional threshold alarms. Historical data is encrypted and stored (AES-256) and supports model retraining (updated every 7 days), reducing the operation and maintenance cost by 30% and extending the equipment life by 10%-15%. By predicting the air flow uniformity index through a BP neural network (η≥0.9), compared with traditional empirical adjustment (η≈0.7), the clean area coverage rate is increased by 20%-30%. Through the linked adjustment of parameters such as temperature and humidity, differential pressure, and wind speed, the compliance rate of the clean workshop with the ISO 14644 standard is increased from 80% to over 98%. Remote real-time monitoring is achieved through the Modbus-TCP / OPC UA protocol, and the firmware and algorithm can be upgraded online, reducing the downtime for maintenance by more than 50%. Through dynamic data fusion, fuzzy PID adaptive control, multi-objective optimization, and intelligent operation and maintenance mechanisms, the present invention systematically solves the defects of traditional clean air-conditioning systems in terms of control accuracy, energy consumption, reliability, and intelligent level. Verified by experiments, it can significantly reduce the operation and maintenance cost, improve the coefficient of performance, and provide a stable and reliable environmental guarantee for high-cleanliness industrial scenarios, with significant economic and social value.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An intelligent clean air conditioning automatic control system, characterized in that: include: Central controller (101), distributed sensor network (102), actuator (103) and control software module (104); The central controller (101) adopts a multi-core processor architecture, has a built-in real-time operating system (RTOS), and is configured with communication interfaces including RS-485, CAN bus and Ethernet; The distributed sensor network (102) is composed of a temperature sensor (accuracy ±0.1°C), a humidity sensor (accuracy ±1%RH), a differential pressure sensor (range 0-500Pa) and a wind speed sensor (range 0-10m / s), which are installed at the four corners and core areas of the clean room, with a sampling frequency of 10Hz, to collect the environmental parameters of the clean room in real time; The actuator (103) includes an electric proportional air valve (A1, opening resolution 0.1%), a variable frequency fan (A2, speed regulation range 10%-100%) and a PID regulating humidifier / dehumidifier (A3, control accuracy ±2%RH), and its opening or speed is adjusted by a control signal output by a central controller (101); The control software module (104) is deployed in the central controller (101), integrates data fusion, fuzzy PID control and multi-objective optimization algorithm, and supports Modbus-TCP protocol remote interaction; The control software module (104) performs weighted processing on the multi-sensor data through a data fusion algorithm, and the weight distribution formula is: Among them, σ i is the measurement standard deviation of the i-th sensor.

2. The intelligent clean air conditioning automatic control system according to claim 1, characterized in that: The central controller (101) adopts a fuzzy PID control algorithm, and its output control quantity u(t) is: Where, e(t) = P target -P real is the environmental parameter deviation, P target is the target environmental parameter of the clean workshop, P real is the actual environmental parameter of the clean room, K p ,K i ,K d is the dynamically adjusted gain factor, K p is the proportional gain, K i is the integral gain), K d is the differential gain, K p (t)∈[0.5,2.0],K i (t)∈[0.01,0.1],K d (t) ∈[0.05,0.5]。 3. The intelligent clean air conditioning automatic control system according to claim 1, characterized in that: The objective function of the multi-objective optimization model of the control software module (104) is: The constraints include: V∈[0.2,0.8]m / s,T∈[20,26]; In the formula, E base is the benchmark energy consumption (unit: kWh), E current is the energy consumption of the current cycle (unit: kWh), P k,range are the cleanroom benchmark environmental parameters, α = 0.6, β = 0.4, and the weights are optimized every 30 minutes using the NSGA-II algorithm.

4. The intelligent clean air conditioning automatic control system according to claim 1, characterized in that: The relationship between the air volume Q and the opening θ of the electric proportional air valve (A1) is: in, ρ=1.2kg / m, opening θ∈[0%,100%], accuracy is guaranteed by 12-bit DAC module.

5. An intelligent clean air conditioning integrated control method, which is implemented based on an intelligent clean air conditioning automatic control system as claimed in any one of claims 1 to 5, characterized in that: The following steps are included: S1. Collect environmental parameters P in real time at a frequency of 10 Hz through a distributed sensor network (102), and eliminate noise through Kalman filtering; S2. Calculate the comprehensive value of weighted environmental parameters based on data fusion algorithm Weight W i Dynamically updated based on sensor health status; S3. Query the fuzzy rule table according to e(t) and γ(t) and output the real-time control parameter K p ,K i ,K d , the fuzzy PID controller is used to generate the control signal U = {u1, u2, u3}, which corresponds to the air valve opening, fan speed and humidification power respectively, and the actuator is dynamically adjusted according to the control signal U; S4. Call the NSGA-II algorithm to optimize α, β every 30 minutes, update the PID parameter threshold, and update the control parameter K through the multi-objective optimization model p ,K i ,K d , and feedback to step S3; S5. Steps S1-S4 are executed in sequence and cyclically, with a control period of 100ms. When an emergency condition is detected (such as a sudden drop in pressure difference exceeding 20%), the interrupt mechanism is triggered, S4 is skipped and the preset safety parameters are directly adopted. Historical data is stored in the edge computing node (105), which supports backtracking analysis and model retraining, and the data retention period is 30 days.

6. The intelligent clean air conditioning integrated control method according to claim 5, characterized in that: The method of collecting environmental parameters P in real time at a frequency of 10 Hz through a distributed sensor network (102) and eliminating noise through Kalman filtering comprises the following steps: The temperature T, humidity H, pressure difference ΔP and wind speed V in the clean room are synchronously collected at a frequency of 10 Hz through a distributed sensor network (102), wherein the temperature sensor adopts a PT100 platinum resistor, the humidity sensor adopts a capacitive polymer film, the pressure difference sensor is a micro-electromechanical system (MEMS) structure, and the wind speed sensor is a hot wire type; The original data is preprocessed by Kalman filtering, and the filtering formula is: In the formula, is the filtering environment parameter at time k, P k is the environmental parameter at time k, is the filtering environment parameter at time k-1, K k is the Kalman gain, H is the observation matrix, the noise covariance matrix Q=0.01I, R=0.05I, where I is the unit matrix; The data is stored in 32-bit floating point format and the transmission integrity is ensured by CRC-16 check.

7. The intelligent clean air conditioning integrated control method according to claim 5, characterized in that: The weighted environmental parameter comprehensive value is calculated based on the data fusion algorithm Weight W i Dynamically update according to the sensor health status, including the following steps: Dynamically adjust the weight W according to the sensor health status i , health status score S i The calculation formula is Among them, R is the number of recent failures; The corrected fusion weight is W i ′=W i ·S i , the comprehensive value is calculated as: If a sensor S i <0.5, the redundant sensor switching is triggered and the fusion calculation logic is updated.

8. The intelligent clean air conditioning integrated control method according to claim 5, characterized in that: The fuzzy rule table is queried according to e(t) and γ(t) to output the real-time control parameter K p ,K i ,K d , using the fuzzy PID controller to generate a control signal U = {u1, u2, u3}, corresponding to the air valve opening, fan speed and humidification power respectively, and dynamically adjusting the actuator according to the control signal U, including the following steps, Fuzzy input variables: environmental parameter deviation e(t) and deviation change rate γ(t), divided into 7 fuzzy sets (NB, NM, NS, ZO, PS, PM, PB); The fuzzy rule table uses the Mamdani reasoning mechanism, and the example rules are: IFe(t)is PB ANDγ(t)is NB THEN K p is PB,K i is NB,K d is PS; Defuzzification uses the centroid method to output accurate PID parameters: where μ j is the activation degree of the jth fuzzy rule; The calculation cycle of the control signal U is 100 ms, and it is output to the actuator (103) through the DA conversion module.

9. The intelligent clean air conditioning integrated control method according to claim 5, characterized in that: The NSGA-II algorithm is called every 30 minutes to optimize α, β, and update the PID parameter threshold, and the control parameter K is updated through the multi-objective optimization model. p ,K i ,K d , and feedback to step S3, including the following steps, The NSGA-II algorithm is used to optimize the objective function weights α and β, with a population size of 50, a crossover probability of 0.8, and a mutation probability of 0.1; Generate a Pareto frontier solution set every 30 minutes, select the optimal solution for energy consumption and parameter deviation, and update the PID parameter threshold: in, is the control parameter threshold before updating, is the updated control parameter threshold, δ is the adaptive learning rate coefficient, which is used to control the amplitude of PID parameter update, and ΔE is the energy consumption optimization rate, which indicates the energy consumption E of the current cycle. current Compared with the benchmark energy consumption E base relative differences; The optimization results are uploaded to the cloud monitoring platform via the OPC UA protocol and a visual report is generated.

10. The intelligent clean air conditioning integrated control method according to claim 5, characterized in that: The steps S1-S4 are executed in sequence and cyclically, with a control period of 100ms; when an emergency condition is detected (such as a sudden drop in pressure difference exceeding 20%), an interruption mechanism is triggered, S4 is skipped and preset safety parameters are directly adopted; historical data is stored in the edge computing node (105), and backtracking analysis and model retraining are supported. The data retention period is 30 days, including the following steps: (a) Periodic control loop execution: Steps S1 to S4 are cyclically executed in sequence, with a control period fixed at 100 ms, and are triggered by a high-precision hardware timer (model STM32F4xx, clock source error ±0.001%) built into the central controller (101); In each cycle, data acquisition takes ≤5ms, data fusion and fuzzy PID calculation takes ≤80ms, and the remaining time is used for asynchronous iterative calculation of multi-objective optimization; (b) Emergency interruption detection and response: Real-time monitoring of the pressure difference parameter ΔP, if |ΔP is satisfied t -ΔP t-1 ∣ / ΔP t-1 If ≥20% lasts for 3 sampling cycles (30ms), it is judged as an emergency condition of sudden pressure drop; Trigger the interrupt service routine (ISR), immediately suspend the current control cycle, skip the optimization step S4, and call the preset safety parameter set U safe ={θ=60%,N=50%,P humid =30%}; In the safety mode, the actuator (103) receives U at a fixed frequency of 10 Hz. safe , until the environmental parameters return to stability (ΔP fluctuation rate < 5% for 10 seconds); (c) Edge data storage and intelligent management: The historical data is stored in an edge computing node (105), which is configured as an ARM Cortex-A72 processor + NVMe SSD, and is written into a time series database (InfluxDB) after being encrypted by timestamp (AES-256). The storage fields include T, H, ΔP, V, U, η and a fault flag, wherein T is the real-time temperature data in the clean room, in degrees Celsius (°C), collected by a temperature sensor, H is the real-time relative humidity data in the clean room, in percentage (%), collected by a humidity sensor (S2), ΔP is the pressure difference data between the clean room and the external environment, in Pascal (Pa), measured by a pressure difference sensor, and used to monitor the air flow direction and airtightness, V is the air flow velocity data in the clean room, in meters per second (m / s), collected by a wind speed sensor, and used to evaluate the air flow organization efficiency, U is a control signal set of the actuator, output by a central controller (101), and used to dynamically adjust the system operation state, and η is an air flow uniformity index; The data retention policy is a rolling deletion mechanism, which automatically cleans up records older than 30 days and generates a compressed archive file (ZIP+CRC32 checksum) and uploads it to the cloud; The model retraining based on the TensorFlow Lite framework is supported. 10% samples (evenly distributed) are extracted from historical data every 7 days to update the LSTM prediction model weights. The optimized model is updated to the central controller (101) via OTA.

Images