An industrial plant low-carbon environment optimization control method and system

Abstract

The present application belongs to the field of multivariable adaptive control system and industrial environment control technology, and relates to an industrial plant low-carbon environment optimization control method and system. The optimization control method comprises: obtaining multiple source physical parameters and operating electrical variables in the plant, and mapping them into a system state vector; constructing a heat and mass diffusion time delay transfer matrix, combining with photovoltaic power fluctuation gradient to calculate compensation time, introducing a dynamic time delay operator into a feedforward loop to generate a feedforward compensation quantity; calculating a dynamic health index and a real-time carbon emission rate, using a continuous S-shaped mapping function to calculate an adaptive weight coefficient, adaptively weighting and arbitrating health and low-carbon oriented feedback quantities, superimposing the feedforward compensation quantity to output a multivariable control vector to drive an actuator to act in advance, and identifying and correcting the feedforward parameters online according to the response residual. The present application has the beneficial effects of offsetting heat and mass transfer delay and eliminating phase lag; weight adaptive arbitration avoids oscillation, and realizes adaptive decoupling of environment quality and low-carbon operation.

Description

Technical Field

[0001] This invention relates to a method and system for optimizing the low-carbon environment of industrial plants, belonging to the fields of multivariable adaptive control systems and industrial environmental control technology. Background Technology

[0002] Environmental control in industrial plants typically employs multi-loop feedback control systems. For parameters such as temperature, humidity, carbon dioxide concentration, and particulate matter concentration, independent control loops are set up. Adjustment commands are generated using proportional-integral-derivative (PID) control algorithms to control the operating frequency of variable frequency fans, the opening of electric fresh air valves, and the opening of air conditioning chilled / heated water valves, thereby achieving closed-loop regulation of various indoor environmental indicators. This type of control method is usually based on the independence of each parameter and the system's ability to respond promptly. However, industrial plants are typically large spaces with strong thermal inertia, and the processes of heat transfer and diffusion in the air are complex. When production conditions change, the number of personnel changes, or the real-time power generation of photovoltaic systems fluctuates, changes in indoor gas concentration and the power consumption of the HVAC system can easily interact, resulting in a relatively long response delay. When increasing feedback gain or lowering the set threshold to accelerate the adjustment speed, the drive mechanism may frequently switch between ventilation and energy-saving frequency limiting, causing control command oscillations, system overshoot, and short-term increases in power consumption. When energy-saving methods such as reducing the operating frequency of variable frequency fans or reducing the number of air exchanges are adopted, it is difficult to respond in time to the sudden diffusion of pollutants, which can easily cause indoor environmental indicators to deviate from the safe range. Even if monitoring nodes are increased or multi-level thresholds are set, it is difficult to eliminate the conflict between different control objectives and the response delay in the heat transfer and mass transfer process due to the lack of feedforward compensation and dynamic weight arbitration mechanism.

[0003] Given the limitations of hardware monitoring and threshold setting, existing multivariate control strategies also struggle to meet the dual demands of large-scale spatial physical delay compensation and dynamic carbon emission management. For example, Chinese invention patent application CN120743014A discloses a substation thermal and humidity environment control method and system based on the Industrial Internet of Things. Although this technical solution acquires parameters through distributed sensors and generates control commands based on a sliding mode control algorithm with a fuzzy exponential approach law, its control mechanism is essentially still a feedback regulation dependent on deviation. In industrial plants with strong thermal inertia and spatial thermal stratification gradients, airflow transmission and heat and mass diffusion have significant physical delays. Pure feedback control cannot respond in advance before significant deviations in state parameters occur, easily leading to control phase lag and overshoot. At the same time, this method does not introduce a dynamic weight arbitration mechanism between environmental health and low-carbon operation, making it difficult to achieve smooth decoupling of multiple control objectives under multi-source energy fluctuations and sudden disturbances.

[0004] Therefore, how to reduce the impact of heat and mass transfer delays on control performance in the high-ceilinged spaces of industrial plants, and achieve coordinated control between environmental quality compliance and low-carbon energy conservation, remains a problem that existing industrial environmental control systems need to solve. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A method for optimizing and controlling the low-carbon environment of industrial plants, comprising:

[0006] Step S1: Obtain multi-source physical parameters and operating electrical variables in the factory by deploying sensor nodes and power monitoring equipment at different heights in the factory, and map them into the system state vector of the current control cycle; the multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at each height; the operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system.

[0007] Step S2: Extract carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculate the spatial gradient change rate, and construct a heat and mass diffusion delay transfer matrix including feedforward compensation parameters. Combine this with real-time photovoltaic power generation to calculate the power fluctuation gradient, and calculate the lead compensation time accordingly. Introduce a dynamic delay operator into the feedforward control loop. Generate feedforward compensation;

[0008] Step S3: Calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient.

[0009] Step S4: Using adaptive weighting coefficients, weighted arbitration is performed on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate. The feedforward compensation quantity is superimposed to output a multivariable control command vector that decouples electric power and environmental regulation.

[0010] Step S5: Send the multivariable control command vector to the terminal regulating mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual.

[0011] Preferably, step S3 specifically includes: calculating the dynamic health index according to the following formula. : ,in, , , These are the carbon dioxide concentration, particulate matter concentration, and temperature collected during the current control cycle; , These are the preset standard carbon dioxide concentration limits and standard particulate matter concentration limits, respectively; For optimal comfort temperature, To allow the maximum temperature deviation; , , The corresponding weight coefficients and The net power consumption is obtained by subtracting the real-time photovoltaic power generation from the real-time operating power of the HVAC system, and then multiplying the net power consumption by the real-time grid carbon emission factor to obtain the real-time carbon emission rate. ,in For the real-time operating power of the HVAC system, Real-time photovoltaic power generation. The real-time grid carbon emission factor is used; the adaptive weighting coefficients are calculated using a continuous sigmoid mapping function, which is expressed as: ,in, For adaptive weighting coefficients, This is the dynamic health index for the current control cycle. To pre-set health warning thresholds, This is the gain adjustment parameter.

[0012] Preferably, step S2 specifically includes: step S21, extracting the carbon dioxide concentration and temperature detected by sensor nodes at different heights within the plant, and calculating the spatial gradient change rate between different heights; step S22, establishing a heat and mass diffusion delay transfer matrix for the plant based on the spatial gradient change rate, and using the difference between the real-time photovoltaic power generation in the current control cycle and the previous control cycle as the power fluctuation gradient; step S23, combining the heat and mass diffusion delay transfer matrix and the power fluctuation gradient, calculating the advance compensation time required for airflow transmission and heat diffusion to each sensor node, and introducing a dynamic delay operator corresponding to the advance compensation time into the feedforward control loop. Generate feedforward compensation amount.

[0013] Preferably, step S4 specifically includes: step S41, calculating the health-oriented feedback quantity based on the dynamic health index, calculating the low-carbon-oriented feedback quantity based on the real-time carbon emission rate, weighting the health-oriented feedback quantity using an adaptive weighting coefficient, and weighting the low-carbon-oriented feedback quantity using the difference between 1 and the adaptive weighting coefficient; step S42, adding the weighted health-oriented feedback quantity and the weighted low-carbon feedback quantity to obtain a composite feedback quantity, and linearly superimposing the composite feedback quantity with the feedforward compensation quantity to generate a multivariable control command vector; the multivariable control command vector includes the variable frequency fan operation frequency control command, the electric fresh air valve opening control command, and the air conditioning cold and hot water source valve opening control command.

[0014] Preferably, step S5 further includes a mechanism for responding to sudden environmental disturbances, specifically including: step S51, constructing a health-low-carbon composite deviation vector based on the deviation between the dynamic health index and the benchmark health target value, and the deviation between the real-time carbon emission rate and the benchmark low-carbon target value, and constructing a positive definite sliding mode switching surface based on the health-low-carbon composite deviation vector; step S52, calculating the real-time distance between the system state trajectory and the positive definite sliding mode switching surface based on the system state trajectory of the health-low-carbon composite deviation vector evolving over time, introducing a sliding mode reaching law and its gain, and dynamically adjusting the sliding mode reaching law gain based on the real-time distance, specifically: increasing the sliding mode reaching law gain when the system state trajectory deviates from the positive definite sliding mode switching surface, and decreasing the sliding mode reaching law gain when the system state trajectory converges to the neighborhood of the positive definite sliding mode switching surface.

[0015] Preferably, step S21 specifically includes: different heights within the plant are divided into the following height levels: near-ground working area, intermediate transition area, and top heat accumulation area, and each height level is equipped with a sensor array of carbon dioxide sensors and temperature sensors; by calculating the temperature difference and carbon dioxide concentration difference between the top heat accumulation area and the near-ground working area, the thermal stratification gradient and gas diffusion gradient of the plant in the vertical direction are obtained, and the spatial gradient change rate per unit time is calculated accordingly, the spatial gradient change rate including the thermal stratification gradient change rate and the gas diffusion gradient change rate.

[0016] Preferably, in step S2, when the real-time photovoltaic power generation is greater than the preset photovoltaic peak power threshold and the power fluctuation gradient is greater than the preset power mutation threshold, it is determined that the photovoltaic power generation is in a peak period and the advanced ventilation and cold storage control is triggered; using the feedforward compensation amount generated during the photovoltaic power generation peak period, the opening of the electric fresh air valve is increased in advance and the opening of the air conditioning cold and heat source water valve is increased, so as to complete the heat storage and cold storage of the plant under the premise that the carbon dioxide concentration is not higher than the standard carbon dioxide concentration limit and the particulate matter concentration is not higher than the standard particulate matter concentration limit.

[0017] Preferably, step S3 further includes: when the carbon dioxide concentration exceeds the preset carbon dioxide limit or the particulate matter concentration exceeds the preset particulate matter limit, the adaptive weighting coefficient is set to a value of 1; the control system releases the weighted constraint of the low-carbon guidance feedback, drives the variable frequency fan and electric fresh air valve at full load, until the dynamic health index falls back to below the preset health warning threshold.

[0018] Preferably, before step S5 is executed, the following steps are included: limiting the frequency control command for the variable frequency fan to a preset upper and lower limit range, and limiting the opening control command for the electric fresh air valve and the opening control command for the air conditioning hot and cold water valve to an opening range of 0% to 100%; when any calculated control command exceeds the corresponding range, clamping and truncating the control command to the corresponding upper or lower limit value.

[0019] A low-carbon environment optimization control system for industrial plants, comprising a method for achieving low-carbon environment optimization control in industrial plants, including:

[0020] The parameter acquisition and vector mapping module is used to acquire multi-source physical parameters and operating electrical variables in the plant through sensor nodes and power monitoring equipment deployed at different heights in the plant, and map them into the system state vector of the current control cycle. The multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at various heights. The operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system.

[0021] The feedforward delay compensation generation module extracts carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculates the spatial gradient change rate, and constructs a heat and mass diffusion delay transfer matrix containing feedforward compensation parameters. It then combines this with real-time photovoltaic power generation to calculate the power fluctuation gradient, thereby calculating the lead compensation time and introducing a dynamic delay operator into the feedforward control loop. Generate feedforward compensation;

[0022] The health and low-carbon arbitration calculation module is used to calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient.

[0023] The composite control vector superposition module is used to use adaptive weighting coefficients to perform weighted arbitration on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate, and superimpose the feedforward compensation quantity to realize the multivariable control command vector decoupling of electric power and environmental regulation.

[0024] The drive control and residual correction module is used to send multivariable control command vectors to the terminal control mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. In the optimization control of low-carbon environment in industrial plants, the control weights of health-oriented feedback and low-carbon-oriented feedback can be continuously adjusted according to the plant's environmental conditions. This avoids the problem of frequent oscillations in the control state when using fixed threshold switching, and allows the control process to smoothly transition between ensuring air quality and reducing energy consumption. This reduces abrupt changes in the operating frequency of variable frequency fans, the opening degree of electric fresh air valves, and the opening degree of air conditioning cold and hot water valves. It also reduces control command oscillations, instantaneous power increases, and overshooting of environmental parameters in the system state vector, thereby improving the stability of the plant's environmental regulation process.

[0027] 2. Based on the temperature and carbon dioxide concentration changes at different heights in the factory building, the delay in airflow transmission and heat diffusion in the high space can be predicted in advance, and the frequency control commands for the operation of the variable frequency fan and the opening commands for the electric fresh air valve can be issued in advance. This can reduce the problem of control actions lagging behind environmental changes. This method can reduce phase lag and control overshoot caused by heat and mass transfer delays, enabling the control system to respond more promptly and smoothly to changes in the factory environment.

[0028] 3. Incorporating real-time photovoltaic power generation changes into the plant environmental control process enables the HVAC system's energy consumption adjustment to be coordinated with the supply of clean energy. When the real-time photovoltaic power generation is high, the system can appropriately advance ventilation, air exchange, and heating / cooling adjustments, thereby reducing the disconnect between environmental control load and photovoltaic output changes, reducing dependence on grid power supply, improving the local utilization rate of photovoltaic power, and enhancing the low-carbon operation effect of the plant environmental control system. Attached Figure Description

[0029] Figure 1 This is a flowchart of the feedforward delay compensation and adaptive arbitration control process of the present invention;

[0030] Figure 2 This is a structural diagram of the vertical layered sensing and terminal driving structure of the factory building according to the present invention.

[0031] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0033] A method for optimizing and controlling the low-carbon environment of an industrial plant includes:

[0034] Step S1: Obtain multi-source physical parameters and operating electrical variables in the factory by deploying sensor nodes and power monitoring equipment at different heights in the factory, and map them into the system state vector of the current control cycle; the multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at each height; the operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system.

[0035] Step S2: Extract carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculate the spatial gradient change rate, and construct a heat and mass diffusion delay transfer matrix including feedforward compensation parameters. Combine this with real-time photovoltaic power generation to calculate the power fluctuation gradient, and calculate the lead compensation time accordingly. Introduce a dynamic delay operator into the feedforward control loop. Generate feedforward compensation;

[0036] Step S3: Calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient.

[0037] Step S4: Using adaptive weighting coefficients, weighted arbitration is performed on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate. The feedforward compensation quantity is superimposed to output a multivariable control command vector that decouples electric power and environmental regulation.

[0038] Step S5: Send the multivariable control command vector to the terminal regulating mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual (using the recursive least squares algorithm).

[0039] Preferably, step S3 specifically includes: calculating the dynamic health index according to the following formula. : in, , , These are the carbon dioxide concentration, particulate matter concentration, and temperature collected during the current control cycle; , These are the preset standard carbon dioxide concentration limits and standard particulate matter concentration limits, respectively; For optimal comfort temperature, To allow the maximum temperature deviation; , , The corresponding weight coefficients and The net power consumption is obtained by subtracting the real-time photovoltaic power generation from the real-time operating power of the HVAC system, and then multiplying the net power consumption by the real-time grid carbon emission factor to obtain the real-time carbon emission rate. ,in For the real-time operating power of the HVAC system, Real-time photovoltaic power generation. For the real-time grid carbon emission factor, adaptive weighting coefficients are calculated using a continuous sigmoid mapping function, which is expressed as follows: ,in, For adaptive weighting coefficients, This is the dynamic health index for the current control cycle. To pre-set health warning thresholds, This is the gain adjustment parameter.

[0040] Preferably, step S2 specifically includes: step S21, extracting the carbon dioxide concentration and temperature detected by sensor nodes at different heights within the plant, and calculating the spatial gradient change rate between different heights; step S22, establishing a heat and mass diffusion delay transfer matrix for the plant based on the spatial gradient change rate, and using the difference between the real-time photovoltaic power generation in the current control cycle and the previous control cycle as the power fluctuation gradient; step S23, combining the heat and mass diffusion delay transfer matrix and the power fluctuation gradient, calculating the advance compensation time required for airflow transmission and heat diffusion to each sensor node, and introducing a dynamic delay operator corresponding to the advance compensation time into the feedforward control loop. Generate feedforward compensation amount.

[0041] Preferably, step S4 specifically includes: step S41, calculating the health-oriented feedback quantity based on the dynamic health index, calculating the low-carbon-oriented feedback quantity based on the real-time carbon emission rate, weighting the health-oriented feedback quantity using an adaptive weighting coefficient, and weighting the low-carbon-oriented feedback quantity using the difference between 1 and the adaptive weighting coefficient; step S42, adding the weighted health-oriented feedback quantity and the weighted low-carbon feedback quantity to obtain a composite feedback quantity, and linearly superimposing the composite feedback quantity with the feedforward compensation quantity to generate a multivariable control command vector; the multivariable control command vector includes a variable frequency fan operation frequency control command, an electric fresh air valve opening control command, and an air conditioning cold and heat source water valve opening control command, which are used to coordinate the ventilation and cold and heat source supply by adjusting electrical variables and physical mechanisms.

[0042] Preferably, step S5 further includes a mechanism to deal with sudden environmental disturbances, specifically including: step S51, constructing a health and low-carbon composite deviation vector based on the deviation between the dynamic health index and the benchmark health target value, and the deviation between the real-time carbon emission rate and the benchmark low-carbon target value, and constructing a positive definite sliding mode switching surface based on the health and low-carbon composite deviation vector; step S52, calculating the real-time distance between the system state trajectory and the positive definite sliding mode switching surface based on the system state trajectory of the health and low-carbon composite deviation vector evolving over time, introducing a sliding mode reaching law and its gain, and dynamically adjusting the sliding mode reaching law gain based on the real-time distance, specifically: increasing the sliding mode reaching law gain to suppress environmental disturbances when the system state trajectory deviates from the positive definite sliding mode switching surface, and decreasing the sliding mode reaching law gain to weaken control command chattering when the system state trajectory converges to the neighborhood of the positive definite sliding mode switching surface.

[0043] Preferably, step S21 specifically includes: different heights within the plant are divided into the following height levels: near-ground working area, intermediate transition area, and top heat accumulation area, and each height level is equipped with a sensor array of carbon dioxide sensors and temperature sensors; by calculating the temperature difference and carbon dioxide concentration difference between the top heat accumulation area and the near-ground working area, the thermal stratification gradient and gas diffusion gradient of the plant in the vertical direction are obtained, and the spatial gradient change rate per unit time is calculated accordingly. The spatial gradient change rate includes the thermal stratification gradient change rate and the gas diffusion gradient change rate, which are the components of the spatial gradient change rate in the vertical direction.

[0044] Preferably, in step S2, when the real-time photovoltaic power generation is greater than the preset photovoltaic peak power threshold and the power fluctuation gradient is greater than the preset power mutation threshold, it is determined that the photovoltaic power generation is in a peak period and the advanced ventilation and cold storage control is triggered; using the feedforward compensation amount generated during the photovoltaic power generation peak period, the opening of the electric fresh air valve is increased in advance and the opening of the air conditioning cold and heat source water valve is increased, so as to complete the heat storage and cold storage of the plant under the premise that the carbon dioxide concentration is not higher than the standard carbon dioxide concentration limit and the particulate matter concentration is not higher than the standard particulate matter concentration limit.

[0045] Preferably, step S3 further includes: when the carbon dioxide concentration exceeds the preset carbon dioxide limit or the particulate matter concentration exceeds the preset particulate matter limit, the adaptive weighting coefficient is set to a value of 1; the control system releases the weighted constraint of the low-carbon guidance feedback, drives the variable frequency fan and electric fresh air valve at full load, until the dynamic health index falls back to below the preset health warning threshold.

[0046] Preferably, before step S5 is executed, the following steps are included: limiting the frequency control command for the variable frequency fan to a preset upper and lower limit range, and limiting the opening control command for the electric fresh air valve and the opening control command for the air conditioning hot and cold water valve to an opening range of 0% to 100%; when any calculated control command exceeds the corresponding range, clamping and truncating the control command to the corresponding upper or lower limit value.

[0047] A low-carbon environment optimization control system for industrial plants, comprising a method for achieving low-carbon environment optimization control in industrial plants, including:

[0048] The parameter acquisition and vector mapping module is used to acquire multi-source physical parameters and operating electrical variables in the plant through sensor nodes and power monitoring equipment deployed at different heights in the plant, and map them into the system state vector of the current control cycle. The multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at various heights. The operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system.

[0049] The feedforward delay compensation generation module extracts carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculates the spatial gradient change rate, and constructs a heat and mass diffusion delay transfer matrix containing feedforward compensation parameters. It then combines this with real-time photovoltaic power generation to calculate the power fluctuation gradient, thereby calculating the lead compensation time and introducing a dynamic delay operator into the feedforward control loop. Generate feedforward compensation;

[0050] The health and low-carbon arbitration calculation module is used to calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient.

[0051] The composite control vector superposition module is used to use adaptive weighting coefficients to perform weighted arbitration on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate, and superimpose the feedforward compensation quantity to realize the multivariable control command vector decoupling of electric power and environmental regulation.

[0052] The drive control and residual correction module is used to send multivariable control command vectors to the terminal control mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual.

[0053] Example 1: This example selects an electronic manufacturing industrial plant with a span of 30m and a net height of 12m. The plant is equipped with a rooftop photovoltaic system and a variable air volume (VAV) HVAC system. The control cycle of the control system is set to 10s, and the standard carbon dioxide concentration limit is... The standard particulate matter concentration limit is 1000 ppm. 75 μg / m³, optimal comfort temperature The maximum allowable temperature deviation is 22℃. The temperature is 5℃, and the weighting coefficient is... , and The values ​​are 0.5, 0.3, and 0.2 respectively. The control system uses the real-time grid carbon emission factor at the current moment in each control cycle. The real-time grid carbon emission factor for this control cycle... The gain adjustment parameter is 0.585 kg CO2 / kWh. Set to 5 to preset the health warning threshold. The value is set to 0.85 to handle the delay in pollutant diffusion caused by changes in personnel density and fluctuations in photovoltaic power, as well as the thermal inertia of the plant.

[0054] When calculating the normalized deviation, the carbon dioxide concentration and particulate matter concentration are calculated by dividing the real-time measured value by the corresponding standard concentration limit. The temperature normalized deviation is calculated by dividing the absolute value of the difference between the temperature and the optimal comfort temperature by the maximum allowable temperature deviation. The preset health warning threshold of 0.85 corresponds to the warning limit when the dynamic health index reaches 0.85. The gain adjustment parameter is determined by fitting the step response curve. When the gain is less than 1, the adaptive weight coefficient changes slowly with the dynamic health index, and the pollutant concentration may exceed the standard concentration limit before the control amount is sufficiently increased. When the gain is greater than 15, the continuous S-shaped mapping function changes too quickly near the preset health warning threshold, which can easily cause control command oscillation. Setting the gain adjustment parameter to 5 allows the adaptive weight coefficient to change continuously in the open interval of 0.10 to 0.95.

[0055] During the control cycle The carbon dioxide concentration collected by the sensor The particulate matter concentration was 1200 ppm. It is 45 μg / m³, at a temperature The temperature was 24.5℃, and the relative humidity was... 55%, real-time photovoltaic power generation The real-time operating power of the HVAC system is 120kW. For a power output of 180kW, the control system maps the aforementioned multi-source physical parameters and operating electrical variables into the system state vector for the current control cycle in a fixed order: , where superscript Representing vector transpose, according to the above method for calculating normalized deviation, the normalized deviation for carbon dioxide is 1.20, for particulate matter it is 0.60, and for temperature it is 0.50. The dynamic health index for the current control cycle is: The net power consumption is 60kW after subtracting the real-time photovoltaic power generation from the real-time operating power of the HVAC system. This net power consumption is then multiplied by the real-time grid carbon emission factor for this control period. To obtain the real-time carbon emission rate It is 35.10 kg CO2 / h.

[0056] The health-oriented feedback and low-carbon-oriented feedback are discretely calculated using a multi-gain proportional-integral controller. The health-oriented feedback is determined by the deviation between the dynamic health index and the benchmark health target value of 0.80, with a health proportional gain coefficient of 40Hz, and the integral cumulative value up to the current control cycle is added. When the dynamic health index is 0.88, the health-oriented feedback is obtained. To increase the frequency by 12 Hz, the low-carbon guidance feedback is determined by the deviation between the real-time carbon emission rate and the baseline low-carbon target value of 26.50 kgCO2 / h, with a low-carbon proportional gain coefficient of 0.58 Hz / (kgCO2 / h). The current real-time carbon emission rate exceeds the baseline low-carbon target value of 8.60 kgCO2 / h, thus yielding the low-carbon guidance feedback. To reduce 5Hz.

[0057] For feedforward compensation, the control system first calculates the spatial gradient between adjacent height levels based on the differences in carbon dioxide concentration and temperature at sensor nodes at different heights. Then, it divides the difference in spatial gradient between the current control cycle and the previous control cycle by the control cycle to obtain the spatial gradient change rate, including the gas diffusion gradient change rate and the thermal stratification gradient change rate. The spatial gradient change rates of each height level are arranged into an input vector according to the sensor node order. The heat and mass diffusion delay transfer matrix is ​​also included. The rows correspond to the sensor nodes to be predicted, and the columns correspond to the gradient rate of change components in the input vector. The transfer terms in the heat and mass diffusion delay transfer matrix adopt a first-order or second-order transfer relationship that includes the inherent transmission delay and inertial elements. The inherent transmission delay, gain parameters, and time constants included in the heat and mass diffusion delay transfer matrix are collectively referred to as feedforward compensation parameters. The initial values ​​of each feedforward compensation parameter are determined based on the step response of the corresponding sensor node. They are used to calculate the response amplitude and phase lag when airflow transmission and heat diffusion reach each sensor node, and are corrected online in real time by the response residuals in subsequent control cycles.

[0058] The difference between the real-time photovoltaic power generation in the current control cycle and the previous control cycle is used as the power fluctuation gradient: The control system multiplies the power fluctuation gradient by a preset energy conversion mapping coefficient to obtain a predicted thermal balance disturbance. This predicted disturbance is then compared with the heat and mass diffusion delay transfer matrix. The calculated phase lag is combined to determine the total time delay required for airflow transmission and heat diffusion to the sensing node. The total time delay is divided by the control cycle and rounded to obtain the number of lead compensation steps. Multiply this number of steps by 10 seconds to obtain the advance compensation time. The controller introduces a dynamic time delay operator corresponding to the lead compensation time in the feedforward control loop. ,according to Each control cycle outputs a feedforward adjustment value in advance, forming a feedforward compensation amount.

[0059] Dynamic health index for this control period The value was 0.88, which is higher than the preset health warning threshold. The adaptive weighting coefficients, at 0.85, are calculated according to a continuous sigmoid mapping function. ,in, The adaptive weighting coefficients for the current control cycle. As the gain adjustment parameter, after substituting the data from the current control cycle, the exponent term is: Therefore, we can conclude that: Based on the calculation results of the heat mass diffusion delay transfer matrix, power fluctuation gradient, and dynamic delay operator, the feedforward compensation amount is converted to the value of the variable frequency fan operating frequency channel. At 0Hz, the control system uses adaptive weighting coefficients to weight the health-oriented and low-carbon-oriented feedback quantities to obtain a composite feedback quantity, which is then linearly superimposed with the feedforward compensation quantity. Based on the reference fan frequency of 50Hz, the variable frequency fan operating frequency control command is as follows: After revision, the operating frequency control command for the variable frequency fan is 54.14Hz, and the opening control command for the electric fresh air valve is adjusted from 40% to 65%. The corresponding control commands are issued in advance according to the advance compensation steps.

[0060] The change in real-time photovoltaic power generation reflects the change in roof solar radiation conditions and alters the photovoltaic power available to the HVAC system. When the real-time photovoltaic power generation decreases, the roof solar radiation heat input changes accordingly, and the clean energy available to the HVAC system also decreases. The control system converts the power fluctuation gradient into the rate of change of the plant's micro-thermal field and the peak-shaving demand of the power grid. Combining the 180s thermal inertia time constant corresponding to airflow transmission and heat diffusion, the system calculates the delay time of the heat and mass balance response through the energy conservation relationship and momentum transfer relationship. The change in photovoltaic power is used as the input to the heat and mass diffusion time delay transfer matrix and feedforward compensation.

[0061] The terminal control mechanism drives the variable frequency fan and electric fresh air valve to operate for three control cycles according to the above control commands. After 30 seconds, the control system collects multi-source physical parameters again to obtain the system state vector. At this time, the carbon dioxide concentration The concentration of particulate matter was reduced to 980 ppm, maintained at 45 μg / m³, and the temperature was maintained at 24.5℃. The dynamic health index was then calculated. It is 0.77, because The exponential term in the continuous sigmoid mapping function is 0.40, and the adaptive weighting coefficients... The number of health-oriented feedback responses decreased to 0.4013. The frequency has been adjusted to increase by 2.98Hz, and the updated variable frequency fan operating frequency control command is as follows: The controller outputs a 48.20Hz variable frequency fan operating frequency control command, and the real-time operating power of the HVAC system drops back to 145kW. The proportion of low-carbon oriented feedback in the composite feedback increases accordingly.

[0062] When dealing with sudden disturbances caused by personnel gatherings or equipment emissions, the control system combines the deviation of the dynamic health index from the benchmark health target value of 0.80, and the deviation of the real-time carbon emission rate from the benchmark low-carbon target value of 26.50 kgCO2 / h, into a health-low-carbon composite deviation vector. Based on this composite deviation vector, a positive definite sliding mode switching surface is constructed: ,in, For sliding mode switching variables, The gain matrix is ​​positive definite. By selecting a positive definite gain matrix, the inner product between the healthy low-carbon composite deviation vector and the sliding mode switching variable is positive, ensuring that the constructed sliding mode switching surface meets the positive definiteness requirement. The control system uses... The modulus is used as the real-time distance between the system state trajectory and the positive definite sliding mode switching surface. When the real-time distance exceeds the preset neighborhood boundary, the sliding mode reaching law gain increases with the distance to accelerate the convergence of the state trajectory to the sliding mode switching surface. After the state trajectory enters the preset neighborhood, the sliding mode reaching law gain decreases with the distance to reduce chattering of the multivariable control command vector.

[0063] When generating and distributing multivariable control command vectors, the controller maps composite control quantities to each terminal regulating mechanism through a preset constant decoupling matrix. The frequency control command for the variable frequency fan operation is directly obtained by weighted superposition of frequency values; the control command for the opening degree of the electric fresh air valve is obtained by multiplying the ratio of the frequency of the variable frequency fan operation to the reference frequency of 50Hz by the reference opening degree of 50%; the control command for the opening degree of the air conditioning chilled and hot water valve is obtained by multiplying the linear combination of the temperature normalization deviation and the humidity normalization deviation by the opening degree gain coefficient of 30%. The resulting multivariable control command vector includes the frequency control command for the variable frequency fan operation, the control command for the opening degree of the electric fresh air valve, and the control command for the opening degree of the air conditioning chilled and hot water valve, and is sent to the corresponding terminal regulating mechanism drive circuits respectively.

[0064] When operating in a multi-region cluster, the control system uses the spatial gradient deviation feedback from sensors in each region to vectorize multivariable control commands. The data is distributed to each terminal control unit. The controller is set with a fixed-duration sliding window containing 100 control cycles. In each control cycle, the predicted state for the next control cycle is calculated using the current heat and mass diffusion delay transfer matrix and the gradient input vector composed of the spatial thermal stratification gradient change rate and the gas diffusion gradient change rate. The sensor acquires the system state vector (actual state) for the next control cycle. Then, the difference between the actual state and the predicted state is used as the response residual. The recursive least squares algorithm uses the gradient input vector as the regression input and the response residual as the observation error to update the heat and mass diffusion delay transfer matrix periodically within the sliding window. The feedforward compensation parameters (including transmission gain and time constant) of each transmission term are updated and used to recalculate the lead compensation time and feedforward compensation amount. This allows the feedforward compensation amount to be adjusted according to model changes caused by thermal resistance, wind resistance and equipment aging inside the plant, and to maintain the continuous output of control commands in each area.

[0065] Example 2: This example involves a comparative test conducted in an electronics manufacturing plant with a span of 30m and a net height of 12m. The plant is divided vertically into a near-ground working area, an intermediate transition area, and a top heat accumulation area. Each area is equipped with a carbon dioxide sensor array with a sampling resolution of 0.1ppm and a temperature sensor array with a measurement accuracy of 0.05℃. Gaussian white noise with a signal-to-noise ratio of 20dB and a 150kW step heat load disturbance are superimposed in the test loop, and the sampling period is controlled. The sampling period is set to 10 seconds. This sampling period is determined based on the time-delay step response of airflow transmission and heat diffusion within the plant, in order to balance control response speed and data processing load.

[0066] The standard carbon dioxide concentration limit used in the test The standard particulate matter concentration limit is 1000 ppm. 75 μg / m³, optimal comfort temperature The maximum allowable temperature deviation is 22℃. The temperature is 5℃, and the weighting coefficient is... , and The real-time grid carbon emission factors for this testing period were 0.5, 0.3, and 0.2, respectively. The gain adjustment parameter is 0.585 kg CO2 / kWh. Set to 5.0, preset health warning threshold Set it to 0.85.

[0067] The experiment was divided into four groups: experimental group, control group 1, control group 2, control group 3, and control group 4. The experimental group adopted a full-process multivariable collaborative control structure that superimposed feedforward delay compensation and adaptive weight arbitration; control group 1 adopted a multi-loop independent PID control method; control group 2 retained adaptive weight arbitration but removed the heat and mass diffusion delay transfer matrix. Corresponding dynamic delay operator Control group three adjusted the gain parameters. Set to 0.5; control group four adjusts the gain parameter. Set to 20.0, except for the differences in the control structure or parameters mentioned above, each group uses the same sensor configuration, disturbance input and initial environmental conditions.

[0068] In the initial operating conditions of the test group, the carbon dioxide concentration was... The particulate matter concentration rose to 1200 ppm. It is 45 μg / m³, at a temperature The temperature was 24.5℃, and the relative humidity was... 55%, real-time photovoltaic power generation The real-time operating power of the HVAC system is 120kW. The system has a power output of 180kW. Each sensor collects the aforementioned multi-source physical parameters and maps them into the system state vector for the current control cycle in a prescribed order. .

[0069] Based on the normalization results of each parameter, the normalization bias for carbon dioxide is 1.20, for particulate matter it is 0.60, and for temperature it is 0.50. The current dynamic health index is obtained by multiplying these three normalization biases by weighting coefficients of 0.5, 0.3, and 0.2 respectively, and then summing them by weight. The value is 0.88. Subtracting the real-time power consumption of the HVAC system from the real-time power generation of the photovoltaic system yields a net power consumption of 60kW. Multiplying the net power consumption by the real-time grid carbon emission factor gives the real-time carbon emission rate. It is 35.10 kg CO2 / h.

[0070] The current dynamic health index of 0.88 is higher than the preset health warning threshold of 0.85. After substituting the deviation between the two into the continuous S-shaped mapping function, the index term is -0.15, and the adaptive weighting coefficient... The value is 0.5374. Based on this, the control system performs weighted arbitration on the health-oriented feedback and the low-carbon-oriented feedback, and superimposes the feedforward compensation generated by the heat and mass diffusion delay transfer matrix and the dynamic delay operator, and outputs a 54.14Hz variable frequency fan operation frequency control command.

[0071] Under the initial operating conditions, the control overshoot of the experimental group was 3.2%, the settling time was 28.5 s, the frequency oscillation amplitude was 0.45 Hz, and the overall carbon reduction rate was 21.4%. As the control command continued to be applied, the carbon dioxide concentration dropped to 980 ppm. According to the system state vector collected under the recovery condition, the dynamic health index dropped to 0.77, and the adaptive weighting coefficient dropped to 0.4013. The frequency control command for the variable frequency fan was adjusted to 48.20 Hz. Under the recovery condition, the control overshoot was 1.1%, the settling time was 12.0 s, the frequency oscillation amplitude was 0.18 Hz, the real-time carbon emission rate was 21.35 kg CO2 / h, and the overall carbon reduction rate was 24.8%.

[0072] The initial carbon dioxide concentration in control group 1 was also 1200 ppm. This group used multi-loop independent PID control, without calculating the dynamic health index and adaptive weight coefficient. After the environmental parameter deviation was detected at the monitoring node, the frequency control command for the variable frequency fan was increased to 60.00 Hz. Since the control action only started to adjust after the pollutants diffused to the monitoring node, the control overshoot of this group was 18.6%, the settling time was 142.0 s, the frequency oscillation amplitude was 8.20 Hz, the real-time carbon emission rate was 50.31 kg CO2 / h, and the overall carbon reduction rate was 0.0%.

[0073] The initial carbon dioxide concentration of control group 2 was 1200 ppm, the dynamic health index was 0.88, the adaptive weighting coefficient was 0.5374, and the frequency control command for the variable frequency fan was 54.14 Hz. After the dynamic delay operator was removed, the frequency control commands for the variable frequency fan, the opening commands for the electric fresh air valve, and the opening commands for the air conditioning chilled and hot water valve could not be issued in advance according to the advance compensation time. The control actions lagged behind the airflow transmission and pollutant diffusion in the high space. The corresponding control overshoot was 11.4%, the settling time was 85.0 s, the frequency oscillation amplitude was 4.10 Hz, the real-time carbon emission rate was 41.20 kg CO2 / h, and the overall carbon reduction rate was 9.2%.

[0074] The initial carbon dioxide concentration of control group 3 was 1200 ppm, the dynamic health index was 0.88, and after the gain adjustment parameter was reduced to 0.5, the change amplitude of the continuous S-shaped mapping function near the preset health warning threshold decreased, the adaptive weighting coefficient was 0.5037, the frequency control command for the variable frequency fan was 50.85 Hz, the weighted proportion of the health-oriented feedback did not increase significantly with the increase of the dynamic health index, and the rate of decrease of carbon dioxide concentration slowed down accordingly. The control overshoot of this group was 14.2%, the adjustment time was 110.0 s, the frequency oscillation amplitude was 3.50 Hz, the real-time carbon emission rate was 38.60 kg CO2 / h, and the comprehensive carbon reduction rate was 13.5%.

[0075] The initial carbon dioxide concentration of control group 4 was 1200 ppm, and the dynamic health index was 0.88. After the gain adjustment parameter was increased to 20.0, the adaptive weighting coefficient recorded by this group was 0.8175, the frequency control command for the variable frequency fan was 59.81 Hz, and the continuous S-shaped mapping function changed rapidly near the preset health warning threshold, causing the frequency control command for the variable frequency fan to repeatedly adjust between high-frequency output and the limited frequency. The control overshoot of this group was 15.8%, the settling time was 95.0 s, the frequency oscillation amplitude was 9.60 Hz, the real-time carbon emission rate was 48.90 kg CO2 / h, and the overall carbon reduction rate was -2.1%.

[0076] During noise disturbances, each group continued to use variable structure correction based on positive definite sliding mode switching surfaces and performed recursive least squares updates using the response residuals within a fixed-duration sliding window. When the dynamic health index exceeded the preset health warning threshold, the experimental group increased the weighting of the health-oriented feedback and issued control commands in advance through feedforward compensation. After the dynamic health index fell back, the adaptive weighting coefficient decreased continuously, and the weighting of the low-carbon-oriented feedback increased accordingly. The comparison results showed that after canceling the dynamic delay operator, the settling time, control overshoot, and frequency oscillation amplitude were all higher than those of the experimental group. When the gain adjustment parameter was too low, the health-oriented feedback increased slowly; when the gain adjustment parameter was too high, the control command generated a large frequency oscillation near the preset health warning threshold.

[0077] Example 3: The current electronic manufacturing industrial plant with a span of 30m and a net height of 12m is divided into a near-ground working area, an intermediate transition area and a top heat accumulation area along the vertical direction. First-level, second-level and third-level sensor arrays are deployed in the three areas respectively. The control system collects the carbon dioxide concentration and temperature of each height level in real time with a control cycle of 10s.

[0078] Within each control cycle, the controller matches sensor nodes at the same horizontal position in different height levels according to their arrangement on the factory floor plan. It calculates the carbon dioxide concentration difference and temperature difference between adjacent height levels, and then divides them by the corresponding vertical height difference to obtain the gas diffusion gradient and thermal stratification gradient at each spatial location. The gradient of the current control cycle is compared with the gradient of the previous control cycle, and the resulting change per unit time is used as the spatial gradient change rate. The input vector is then formed according to the spatial arrangement order of the sensor nodes.

[0079] The controller establishes a heat and mass diffusion delay transfer matrix based on the spatial gradient change rate. The transfer terms in the matrix adopt a first- or second-order frequency domain transfer relationship that includes inherent transmission delay and inertial elements. The gain parameter is set to 0.8 to 1.5, and the time constant is set to 120s to 180s. It is used to calculate the response amplitude and phase lag of airflow transmission and heat diffusion at each sensing node.

[0080] The difference between the real-time photovoltaic power generation in the current control cycle and the previous control cycle is used as the power fluctuation gradient. The preset power change threshold is converted into the power difference corresponding to the adjacent control cycle according to the photovoltaic power change rate of 10kW / min. When the power fluctuation gradient exceeds the threshold, the controller triggers the advance compensation calculation. During the calculation, the power fluctuation amplitude is multiplied by a fixed energy conversion mapping coefficient to obtain the predicted thermal balance disturbance. This is then combined with the gas transmission phase lag calculated by the heat and mass diffusion delay transfer matrix to determine the total time delay required for airflow transmission and heat diffusion to the sensing node.

[0081] The controller divides the total delay by a 10-second control cycle and rounds down to obtain the number of lead compensation steps. In this embodiment, the calculated number of lead compensation steps is 2 control cycles, corresponding to a lead compensation time of... The time delay is 20 seconds, and the feedforward control loop uses a dynamic time delay operator. The feedforward adjustment value is processed with a time delay to generate a feedforward compensation amount, so that the frequency control command for the variable frequency fan and the opening control command for the electric fresh air valve are issued two control cycles in advance, in order to offset the response lag caused by airflow transmission and pollutant diffusion in high spaces.

[0082] When personnel enter the factory or production equipment suddenly emits gaseous pollutants, the control system generates a health-low-carbon composite deviation vector based on the deviation of the dynamic health index from the benchmark health target value and the deviation of the real-time carbon emission rate from the benchmark low-carbon target value. The dynamic health index is obtained by weighting the normalized biases of carbon dioxide concentration, particulate matter concentration and temperature, with the weights of the three biases being 0.50, 0.30 and 0.20, respectively.

[0083] The controller is based on the healthy and low-carbon composite deviation vector. Constructing positive definite sliding mode switching surfaces The magnitude of the sliding mode switching variable is used as the real-time distance between the system state trajectory and the positive definite sliding mode switching surface. When the real-time distance is greater than the preset boundary limit, the sliding mode reaching law gain is increased to make the state trajectory converge toward the positive definite sliding mode switching surface. When the state trajectory enters the preset neighborhood, the sliding mode reaching law gain is reduced as the real-time distance decreases, and a continuous saturated function operator is used to replace the symbolic step function, so that the frequency control command of the variable frequency fan, the opening control command of the electric fresh air valve, and the opening control command of the air conditioning cold and hot water valve change continuously, reducing speed regulation chatter. During this control process, the carbon dioxide concentration is kept within the standard carbon dioxide concentration limit of 1000 ppm.

[0084] When constructing the dynamic health index, the controller first normalizes the environmental parameters of different dimensions: the carbon dioxide concentration deviation is divided by the standard carbon dioxide concentration limit of 1000 ppm, the particulate matter concentration deviation is divided by the standard particulate matter concentration limit of 75 μg / m³, and the temperature deviation is divided by the maximum allowable temperature deviation of 5℃. The three dimensionless deviations are then weighted and summed according to 0.50, 0.30, and 0.20, respectively, to obtain the dynamic health index for the current control cycle.

[0085] The control system outputs a multivariable control command vector in each control cycle. It drives the variable frequency fan and electric fresh air valve to operate, and collects the system state vector (actual state) in the next control cycle. The controller uses the current heat mass diffusion delay transfer matrix and the spatial gradient change rate of the current cycle to calculate the predicted state for the next control cycle. Then, the difference between the actual state and the predicted state is used as the response residual.

[0086] Online identification employs a fixed-duration sliding window containing 100 control cycles. The recursive least squares algorithm uses the input vector composed of the spatial gradient change rate as the regression input and the response residual as the observation error. After discretizing the continuous frequency domain transfer relationship through bilinear transformation, the feedforward compensation parameters (including transmission gain and time constant) in the heat and mass diffusion delay transfer matrix are corrected cycle by cycle. The sampled residual vector is a vector composed of the difference between the actual state and the predicted state of each height-level sensor node in the next control cycle in each component. The sampled residual vector is projected according to each regression component to form the corresponding scalar residual, which is used to calculate the parameter correction amount for the current control cycle.

[0087] When the mechanical ventilation resistance inside the plant changes, or when dust accumulation on the filter increases the air resistance, the response residual between the actual state and the predicted state increases accordingly. The recursive least squares algorithm updates the heat and mass diffusion delay transfer matrix based on this response residual. The feedforward compensation parameters (including transmission gain and time constant) are calculated, and the lead compensation time is recalculated. The updated dynamic delay operator adjusts the advance issuance steps of the variable frequency fan operation frequency control command and the electric fresh air valve opening control command accordingly, so that the multivariable control command vector formed by the weighted superposition of the feedforward compensation, health-oriented feedback and low-carbon-oriented feedback is consistent with the actual airflow transmission and heat diffusion state in the plant.

[0088] Example 4: This example combines Figures 1 to 2 This paper describes a method and system for optimizing and controlling the low-carbon environment of an industrial plant, such as... Figure 1As shown, in step S1, multi-source physical parameters and operating electrical variables within the plant are acquired by sensor nodes and power monitoring equipment deployed at different heights within the plant, and mapped to the system state vector of the current control cycle. In step S2, the spatial gradient change rate is extracted to construct a heat and mass diffusion delay transfer matrix containing feedforward compensation parameters. The advance compensation time is calculated by combining the power fluctuation gradient, and a dynamic delay operator is introduced to generate the feedforward compensation quantity. In step S3, the dynamic health index and real-time carbon emission rate are calculated based on the system state vector, and the adaptive weighting coefficient is calculated using a continuous S-shaped mapping function. In step S4, the health-oriented and low-carbon-oriented feedback quantities are weighted and arbitrated using the adaptive weighting coefficient, and the feedforward compensation quantity is superimposed to output a multivariable control command vector. In step S5, the multivariable control command vector is sent to the terminal adjustment mechanism, the system state vector of the next control cycle is collected as the actual state, the predicted state of the next control cycle is calculated based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, the response residual between the actual state and the predicted state is calculated, and the feedforward compensation parameters are corrected online.

[0089] like Figure 2 As shown, the interior of the electronic manufacturing plant is divided into a top heat accumulation zone, a middle transition zone, and a near-ground working area. A third-level sensor array is deployed in the top heat accumulation zone, a second-level sensor array in the middle transition zone, and a first-level sensor array in the near-ground working area. These three sensor arrays collect and output multi-source physical parameters to a digital controller. The rooftop photovoltaic system outputs real-time photovoltaic power to the digital controller. The digital controller contains a control motherboard. It receives the multi-source physical parameters and the real-time photovoltaic power and outputs multi-variable control command vectors to the drive circuit. The drive circuit receives these multi-variable control command vectors and sends control commands to downstream terminal adjustment mechanisms. Specifically, it sends control commands to the air conditioning cold and hot water valves to control their opening degree, to the electric fresh air valves to control their opening degree, and to the variable frequency fans to control their operating frequency.

[0090] Example 5: Before the industrial plant environmental control system is put into operation, the controller uses the shutdown period when there is no personnel disturbance and no production heat to complete the zero-point calibration of the sensors and the initialization of the feedforward compensation parameters. During calibration, the electric fresh air valve is completely closed, and the variable frequency fan is driven to run at a fixed frequency for 30 minutes to continuously collect the electrical signal readings of the first, second and third level sensor arrays. After obtaining 180 sampling points for each sensor node, its average value is calculated and written into the controller register as the initial zero-point reference value of the corresponding node, which is used to correct the sensor zero-point drift caused by electrode aging and changes in transmission cable impedance.

[0091] After completing the zero-point calibration, the controller sends a step frequency command to the variable frequency fan, abruptly changing from 35Hz to 50Hz. Simultaneously, it collects the dynamic response waveforms of carbon dioxide concentration and temperature from the top heat accumulation zone to the near-ground working zone. The controller takes the time difference between the start of the response curve and the inflection point of the peak rate of change as the heat and mass diffusion delay transfer matrix. The initial delay parameter is used to calculate the initial lead compensation time. The initialized feedforward compensation parameters are then fixed on the control motherboard.

[0092] In digital controllers, dynamic time delay operators in the continuous frequency domain This is implemented using a discrete-time circular first-in-first-out data buffer queue, with the controller providing advance compensation time. Divide by the 10s control cycle to obtain the number of advance compensation steps in integer form; within each control cycle, write the currently generated feedforward adjustment value to the end of the buffer queue, and make the output pointer read the feedforward adjustment value at the corresponding position in advance according to the number of steps, thereby forming the feedforward compensation amount issued in advance.

[0093] When seasonal operating conditions change or external weather conditions change significantly in the plant, the control system periodically performs sensor node calibration and fault isolation. After each 24-hour operating cycle, the controller calculates the sampling variance of carbon dioxide concentration, particulate matter concentration, and temperature collected by each sensor node. When the sampling variance of a single sensor node exceeds a preset noise envelope threshold, the output signal of that node is blocked, and the measured value of that node is replaced by the spatially weighted average of adjacent sensor nodes at the same height level, and the system state vector for the current control cycle is reconstructed. .

[0094] If the dynamic health index is not triggered when the limit concentration protection is not activated. When upgraded to version 1.0 or higher, the controller will output adaptive weighting coefficients from the continuous sigmoid mapping function. The values ​​are limited to an open range of 0.10 to 0.95, so that the health-oriented feedback quantity and the low-carbon-oriented feedback quantity are continuously weighted. When converting the concentration recovery judgment value from the preset health warning threshold, the threshold is multiplied by the standard carbon dioxide concentration limit and the standard particulate matter concentration limit respectively to obtain the carbon dioxide concentration recovery judgment value and the particulate matter concentration recovery judgment value.

[0095] When the carbon dioxide concentration exceeds the corresponding preset limit concentration limit, or the particulate matter concentration exceeds the corresponding preset limit concentration limit, the limit concentration protection takes precedence over the conventional limiting rule with adaptive weighting coefficient, and the controller will... Setting it directly to 1 removes the weighted constraint of the low-carbon-oriented feedback, and drives the variable frequency fan and electric fresh air valve at full load. The controller continuously collects the concentrations of the two types of pollutants until the carbon dioxide concentration is lower than the carbon dioxide concentration recovery judgment value and the particulate matter concentration is lower than the particulate matter concentration recovery judgment value. Then, the continuous S-shaped mapping function is restored to the weighted arbitration of the health-oriented feedback and the low-carbon-oriented feedback, so that the multivariable control command vector... Continuous output to the variable frequency fan and electric fresh air valve.

[0096] Example 6: This example sets up four independent processing zones, each zone being divided into an independent control unit. Each control unit is equipped with a carbon dioxide sensor with a sampling resolution of 0.1 ppm, a temperature sensor with a measurement accuracy of 0.05℃, and an airflow control actuator. It collects carbon dioxide concentration, particulate matter concentration, temperature, relative humidity, real-time photovoltaic power generation, and real-time HVAC system operating power in a 10-second control cycle, and maps these data to the [data missing]. Cross-regional control cycle System state vector: , where subscript The cross-region number is assigned a value ranging from 1 to 4. The control system calculates the dynamic health index based on the system state vector of each cross-region. And utilize it in conjunction with preset health warning thresholds The deviation between them is calculated using a continuous S-shaped mapping function to determine the adaptive weighting coefficients for each cross-region. .

[0097] The total grid power limit of the factory building and the real-time photovoltaic power generation Under the constraints, the control system adjusts the carbon dioxide concentration change rate and dynamic health index of each cross-region. Determine the resource allocation ratio, and then redistribute the total fresh air volume and heating / cooling source supply to achieve a dynamic health index. Exceeding the preset health warning threshold Cross-regional areas are given priority in receiving higher frequency control commands for variable frequency fan operation and larger fresh air supply; for other cross-regional areas where the dynamic health index does not exceed the preset health warning threshold and the air quality meets the standards, the frequency control commands for variable frequency fan operation are reduced accordingly, so that the environmental regulation needs of each cross-regional area match the available power of the plant.

[0098] When lifting equipment is moved or temporary partitions are placed inside the factory, the heat and mass transfer path and wind resistance distribution change along with the spatial physical topology. The control system compares the changes cycle by cycle within a sliding window of 60 consecutive control cycles. actual status of sensors across regions Compared with the predicted state calculated based on the pre-corrected heat mass diffusion delay transfer matrix The prediction mean square residual is calculated after converting the response residuals of each state component into relative deviations.

[0099] When the predicted mean square residual exceeds a preset 5% deviation threshold, the control system triggers the parameter identification procedure, employing a recursive least squares algorithm to identify the first parameter. Cross-regional heat and mass diffusion delay transfer matrix The feedforward compensation parameters (including gain parameters and time constants) are corrected online, and the corrected parameters are used to recalculate the lead compensation time for this span. And update the dynamic delay operator in the feedforward control loop. .

[0100] Updated dynamic delay operator Based on the recalculated advance compensation time, the advance issuance steps of the variable frequency fan operation frequency control command and the electric fresh air valve opening control command are adjusted to compensate for the additional mass transfer delay caused by changes in spatial physical topology, so that the carbon dioxide concentration in each cross-zone is restored to and maintained within the standard carbon dioxide concentration limit. The corresponding value is within 1000 ppm.

[0101] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for optimizing and controlling the low-carbon environment of an industrial plant, characterized in that, include: Step S1: Obtain multi-source physical parameters and operating electrical variables in the factory by deploying sensor nodes and power monitoring equipment at different heights in the factory, and map them into the system state vector of the current control cycle; the multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at each height; the operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system. Step S2: Extract carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculate the spatial gradient change rate, and construct a heat and mass diffusion delay transfer matrix including feedforward compensation parameters. Combine this with real-time photovoltaic power generation to calculate the power fluctuation gradient, and calculate the lead compensation time accordingly. Introduce a dynamic delay operator into the feedforward control loop. Generate feedforward compensation; Step S3: Calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient. Step S4: Using adaptive weighting coefficients, weighted arbitration is performed on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate. The feedforward compensation quantity is superimposed to output a multivariable control command vector that decouples electric power and environmental regulation. Step S5: Send the multivariable control command vector to the terminal regulating mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual.

2. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 1, characterized in that, Step S3 specifically includes: calculating the dynamic health index according to the following formula. : ,in, , , These are the carbon dioxide concentration, particulate matter concentration, and temperature collected during the current control cycle; , These are the preset standard carbon dioxide concentration limits and standard particulate matter concentration limits, respectively; For optimal comfort temperature, To allow the maximum temperature deviation; , , The corresponding weight coefficients and The net power consumption is obtained by subtracting the real-time photovoltaic power generation from the real-time operating power of the HVAC system, and then multiplying the net power consumption by the real-time grid carbon emission factor to obtain the real-time carbon emission rate. ,in For the real-time operating power of the HVAC system, Real-time photovoltaic power generation. The real-time grid carbon emission factor is used; the adaptive weighting coefficients are calculated using a continuous sigmoid mapping function, which is expressed as: ,in, For adaptive weighting coefficients, This is the dynamic health index for the current control cycle. To pre-set health warning thresholds, This is the gain adjustment parameter.

3. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 1, characterized in that, Step S2 specifically includes: Step S21, extracting the carbon dioxide concentration and temperature detected by sensor nodes at different heights within the plant, and calculating the spatial gradient change rate between different heights; Step S22, establishing a heat and mass diffusion delay transfer matrix for the plant based on the spatial gradient change rate, and using the difference between the real-time photovoltaic power generation in the current control cycle and the previous control cycle as the power fluctuation gradient; Step S23, combining the heat and mass diffusion delay transfer matrix and the power fluctuation gradient, calculating the advance compensation time required for airflow transmission and heat diffusion to each sensor node, and introducing a dynamic delay operator corresponding to the advance compensation time into the feedforward control loop. Generate feedforward compensation amount.

4. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 1, characterized in that, Step S4 specifically includes: Step S41, calculating the health-oriented feedback quantity based on the dynamic health index, calculating the low-carbon-oriented feedback quantity based on the real-time carbon emission rate, weighting the health-oriented feedback quantity using an adaptive weighting coefficient, and weighting the low-carbon-oriented feedback quantity using the difference between 1 and the adaptive weighting coefficient; Step S42, adding the weighted health-oriented feedback quantity and the weighted low-carbon feedback quantity to obtain a composite feedback quantity, and linearly superimposing the composite feedback quantity with the feedforward compensation quantity to generate a multivariable control command vector; the multivariable control command vector includes the variable frequency fan operation frequency control command, the electric fresh air valve opening control command, and the air conditioning cold and hot water source valve opening control command.

5. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 1, characterized in that, Step S5 also includes a mechanism to deal with sudden environmental disturbances, specifically including: Step S51, constructing a health and low-carbon composite deviation vector based on the deviation between the dynamic health index and the benchmark health target value, and the deviation between the real-time carbon emission rate and the benchmark low-carbon target value, and constructing a positive definite sliding mode switching surface based on the health and low-carbon composite deviation vector; Step S52, calculating the real-time distance between the system state trajectory and the positive definite sliding mode switching surface based on the system state trajectory of the health and low-carbon composite deviation vector evolving over time, introducing the sliding mode reaching law and its gain, and dynamically adjusting the sliding mode reaching law gain based on the real-time distance, specifically: increasing the sliding mode reaching law gain when the system state trajectory deviates from the positive definite sliding mode switching surface, and decreasing the sliding mode reaching law gain when the system state trajectory converges to the neighborhood of the positive definite sliding mode switching surface.

6. The method for optimizing and controlling a low-carbon environment in an industrial plant according to claim 3, characterized in that, Step S21 specifically includes: different heights within the plant are divided into the following height levels: near-ground working area, intermediate transition area, and top heat accumulation area. Each height level is equipped with a sensor array consisting of carbon dioxide sensors and temperature sensors. By calculating the temperature difference and carbon dioxide concentration difference between the top heat accumulation area and the near-ground working area, the thermal stratification gradient and gas diffusion gradient of the plant in the vertical direction are obtained. Based on this, the spatial gradient change rate per unit time is calculated, which includes the thermal stratification gradient change rate and the gas diffusion gradient change rate.

7. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 1, characterized in that, In step S2, when the real-time photovoltaic power generation is greater than the preset photovoltaic peak power threshold and the power fluctuation gradient is greater than the preset power mutation threshold, it is determined that the photovoltaic power generation is in a peak period and the advanced ventilation and cold storage control is triggered. Using the feedforward compensation generated during the photovoltaic power generation peak period, the opening of the electric fresh air valve is increased in advance and the opening of the air conditioning cold and heat source water valve is increased. Under the premise of keeping the carbon dioxide concentration not higher than the standard carbon dioxide concentration limit and the particulate matter concentration not higher than the standard particulate matter concentration limit, the heat storage and cold storage of the plant are completed.

8. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 2, characterized in that, Step S3 also includes that when the carbon dioxide concentration exceeds the preset carbon dioxide limit or the particulate matter concentration exceeds the preset particulate matter limit, the adaptive weighting coefficient is set to a value of 1; the control system releases the weighted constraint of the low-carbon guidance feedback, drives the variable frequency fan and electric fresh air valve at full load, until the dynamic health index falls back to below the preset health warning threshold.

9. The method for optimizing and controlling the low-carbon environment of an industrial plant according to claim 4, characterized in that, Before step S5 is executed, the following steps are taken: limiting the frequency control command for the variable frequency fan to the upper and lower limits of a preset frequency range, and limiting the opening control command for the electric fresh air valve and the opening control command for the air conditioning cold and hot water valve to the opening range of 0% to 100%; when any calculated control command exceeds the corresponding range, clamping and truncating the control command to the corresponding upper or lower limit value.

10. A low-carbon environment optimization control system for industrial plants, used to implement the low-carbon environment optimization control method for industrial plants as described in claim 1, characterized in that, include: The parameter acquisition and vector mapping module is used to acquire multi-source physical parameters and operating electrical variables in the plant through sensor nodes and power monitoring equipment deployed at different heights in the plant, and map them into the system state vector of the current control cycle. The multi-source physical parameters include carbon dioxide concentration and temperature, as well as particulate matter concentration and relative humidity detected by sensor nodes at various heights. The operating electrical variables include real-time photovoltaic power generation and real-time operating electrical power of the HVAC system. The feedforward delay compensation generation module extracts carbon dioxide concentration and temperature data detected by sensor nodes at different heights within the plant, calculates the spatial gradient change rate, and constructs a heat and mass diffusion delay transfer matrix containing feedforward compensation parameters. It then combines this with real-time photovoltaic power generation to calculate the power fluctuation gradient, thereby calculating the lead compensation time and introducing a dynamic delay operator into the feedforward control loop. Generate feedforward compensation; The health and low-carbon arbitration calculation module is used to calculate the dynamic health index and real-time carbon emission rate of the current control cycle based on the system state vector, calculate the deviation between the dynamic health index and the preset health warning threshold, and input the deviation as an independent variable into the continuous S-shaped mapping function to calculate the adaptive weight coefficient. The composite control vector superposition module is used to use adaptive weighting coefficients to perform weighted arbitration on the health-oriented feedback quantity generated based on the dynamic health index and the low-carbon-oriented feedback quantity generated based on the real-time carbon emission rate, and superimpose the feedforward compensation quantity to realize the multivariable control command vector decoupling of electric power and environmental regulation. The drive control and residual correction module is used to send multivariable control command vectors to the terminal control mechanism, collect the system state vector of the next control cycle as the actual state, calculate the predicted state of the next control cycle based on the heat and mass diffusion delay transfer matrix and the system state vector of the current control cycle, calculate the response residual between the actual state and the predicted state, and correct the feedforward compensation parameters of the heat and mass diffusion delay transfer matrix online based on the response residual.

Images