Energy-saving control equipment applied to zeolite runner system

Abstract

The invention discloses zeolite runner system energy-saving control equipment applied to VOCs waste gas treatment, and belongs to the field of environmental protection and energy conservation. The system aims at solving the problems that a traditional system depends on a fixed control rule and is difficult to adapt to VOCs concentration and waste gas flow fluctuation, so that energy consumption is high, and efficiency is unstable. The device core comprises a device body, an integrated AI algorithm controller and a human-computer interaction interface. The AI controller is internally provided with a BP neural network, reinforcement learning and an MPC model, receives signals of VOCs concentration, flow, desorption temperature, rotating wheel rotating speed and the like, carries out multi-algorithm collaboration-BP fitting nonlinear correlation, corrects fluctuation deviation through reinforcement learning, ensures that parameters are compliant through MPC, and carries out real-time monitoring on VOCs. And rotating speed and desorption temperature adjusting instructions are output to drive equipment. The human-computer interface is a 7-inch capacitive screen, displays data in a partition mode, supports touch forbidding and anti-misoperation, and interacts with the AI controller through RS485 (Modbus-RTU). The equipment adapts to working condition fluctuation, energy consumption is reduced, and operation stability and operation safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy-saving technology for environmental protection equipment, in particular to a zeolite rotary system energy-saving control device applied to volatile organic compounds (VOCs) waste gas treatment, which is suitable for industrial scenes such as automobiles, ships, painting, printing and chemical industry, and realizes energy consumption optimization through dynamic adjustment of rotary operation parameters, belonging to the cross field of VOCs treatment and energy-saving control. BACKGROUND

[0002] The zeolite rotary system is the mainstream equipment for industrial treatment of low-concentration and large-volume VOCs waste gas, and the core structure includes a pretreatment unit (to filter dust and particulate matter), a rotary main body (divided into an adsorption zone, a desorption zone and a cooling zone), a heating unit (to supply desorption hot air) and a fan unit (to drive air circulation). The operation principle is as follows: the pretreated VOCs waste gas enters the adsorption zone, is separated and purified by zeolite molecular sieve adsorption, and the purified gas is discharged in compliance with the standard; the saturated zeolite rotates to the desorption zone with the rotary, is desorbed and regenerated under 180-220℃ hot air, and is cooled in the cooling zone to return to the adsorption zone, forming a continuous closed loop of "adsorption-desorption-cooling".

[0003] Although the system has high VOCs treatment efficiency, the traditional design and control have significant energy-saving defects, and the operation energy consumption is high (accounting for 40%-60% of the total energy consumption of enterprise environmental protection equipment), which is difficult to adapt to the "double carbon" demand, and the specific reasons are as follows:

[0004] Parameter fixation causes energy redundancy: the core parameters (such as rotary speed 2-4r / h, desorption temperature 180-220℃, and fan full load) are fixed according to "maximum design working condition" (such as VOCs concentration 1500mg / m 3 , waste gas flow 10000m 3 / h), which cannot be adjusted with working condition. For example, when the VOCs concentration decreases from 1000mg / m 3 to 200mg / m 3 during intermittent production, the system still maintains high parameter operation, the heating unit has invalid energy consumption of 40%-60%, and the fan has redundant energy consumption of 30%-50%, increasing the cost of enterprises.

[0005] No dynamic response makes it difficult to balance efficiency and energy saving: the VOCs concentration and flow of industrial waste gas fluctuate with production (such as 50%-200% fluctuation of painting color concentration and 60% reduction of intermittent production flow), but the traditional system cannot adjust the parameters in real time: maintaining high parameters causes energy waste at low load, and reducing parameters easily causes incomplete desorption or adsorption saturation, resulting in decreased purification efficiency and emission exceeding the standard (outlet VOCs>30mg / m 3 ).

[0006] Fixed working condition acceleration wear increases cost: the system runs at the maximum speed (2-4 r / h) for a long time, and the wear rate of the runner driving part increases by 30%-50%; the continuous high temperature (above 200℃) accelerates the aging of zeolite (the rate increases by 20%-40%), and the service life is reduced from 5 years to 3-4 years. Frequent maintenance (2-3 times more per year) and adsorbent replacement (more than 100,000 yuan per time) not only increase operation and maintenance costs, but also reduce energy consumption efficiency due to downtime. SUMMARY

[0007] The present application aims to solve the problems of high energy consumption, difficult balance between treatment efficiency and energy saving, and fast equipment wear caused by parameter fixation and simple control method of traditional zeolite runner system, and provides an energy-saving control device applied to a zeolite runner system, which realizes minimization of energy consumption under the premise of ensuring waste gas treatment efficiency (≥90%) by accurately detecting waste gas working conditions, intelligently dynamically adjusting operating parameters and forming a closed-loop control.

[0008] To achieve the above-mentioned purpose, the present application provides an energy-saving control device applied to a zeolite runner system, and the core technical scheme is as follows:

[0009] The device includes a gas pretreatment unit, a concentration detection unit, a state feedback unit, a control unit, and a matching communication, alarm and furnace temperature control unit, which work cooperatively to realize dynamic energy-saving control by interacting with the PLC controller of the zeolite runner system, and the specific structure and functions are as follows:

[0010] The gas pretreatment unit is used for pretreating the waste gas entering the detection link to ensure the accuracy of subsequent concentration detection, and includes a filter 1, a suction pump, a filter 2, a condenser and a peristaltic pump connected in series:

[0011] The filter 1 and the filter 2 are both high molecular filters made of PE material, which can effectively intercept impurities (such as dust and particulate matter) in the waste gas (filtration efficiency ≥90%) to avoid the impurities adhering to the surface of the detection element to affect the precision; the filter 1 is used for initially intercepting particulate impurities in the waste gas, the suction pump is used for extracting the industrial waste gas filtered by the filter 1 and driving the system air circulation, and the filter 2 is used for secondarily intercepting residual particulate impurities in the waste gas; the condenser reduces the humidity of the waste gas by temperature control (the condensation temperature is set to 5-10℃) to stabilize the relative humidity of the waste gas to ≤30%, thereby preventing the interference of high humidity on the detection of VOCs concentration (such as the influence of water on ionization efficiency in the PID detection method). The peristaltic pump is used for stably pumping the sewage generated in the filtration process and the liquid impurities condensed in the waste gas, and the pumping flow range is 0.1-2 L / h, and the peristaltic pump has anti-blocking and anti-stalling protection functions.

[0012] The concentration detection unit is used for real-time acquisition and output of the core parameters of the waste gas to provide a data basis for working condition judgment and instruction generation, and includes a PID concentration meter and a flow meter:

[0013] The flow meter adopts a rotor flow meter to stably and adjustably control the extraction flow of the exhaust gas sample in the range of 0.5-5 m 3 / h, so as to avoid flow fluctuation to cause inconsistent detection conditions;

[0014] The PID concentration instrument is used to detect the concentration value of the target pollutant (such as non-methane total hydrocarbon) in the exhaust gas sample in real time, has a detection accuracy of ±5% FS (full scale), a response time of ≤10 s, and can quickly output a concentration signal; at the same time, the flow meter synchronously outputs a real-time flow signal, and the two are collectively used as core input data for working condition analysis.

[0015] The state feedback unit is connected with the PLC controller signal of the zeolite rotary system, and is used to receive the actual rotary speed signal and the actual desorption temperature signal of the zeolite rotary system transmitted by the PLC controller in real time;

[0016] The actual rotary speed signal and the actual desorption temperature signal come from:

[0017] The rotary speed sensor is installed at the output end of the zeolite rotary drive motor, and is used to collect the actual rotary speed signal (unit: r / h);

[0018] The desorption temperature sensor is inserted into the hot air pipeline in the resolution zone of the zeolite rotary, and is used to collect the actual desorption temperature signal (unit: ℃);

[0019] The control unit is a core regulation module of the device, and contains an AI algorithm controller, which realizes the intelligent regulation process of “detection-analysis-decision-instruction output-feedback optimization” by interacting with the PLC controller of the zeolite rotary system:

[0020] The AI algorithm controller is connected with the concentration detection unit and the state feedback unit, and is used to receive the concentration signal, the flow signal, the actual rotary speed signal and the actual desorption temperature signal; the built-in machine learning algorithm based on the BP neural network takes the historical operation data (including the exhaust gas concentration, the flow, the rotary speed, the desorption temperature and the corresponding energy consumption data) as the training sample, and takes the minimization of the device energy consumption when the exhaust gas treatment efficiency is ≥90% as the optimization target; through real-time calculation, the optimal rotary speed adjustment instruction and the desorption temperature adjustment instruction corresponding to the current exhaust gas working condition are generated, and the instructions are transmitted to the PLC controller of the zeolite rotary system.

[0021] The structure of the BP neural network algorithm is that the input layer contains 4 neurons (corresponding to the concentration value, the flow value, the actual rotary speed and the actual desorption temperature), the output layer contains 2 neurons (corresponding to the rotary speed adjustment value and the desorption temperature adjustment value), and the hidden layer is set to 2 layers (12 neurons in each layer), so as to ensure the accuracy of the adjustment instruction and avoid excessive regulation or insufficient regulation.

[0022] Closed-loop control and extended functions: The zeolite wheel system's built-in PLC controller receives the adjustment instructions output by the AI algorithm controller, controls the drive motor and desorption heating device, and executes the action of the actuator. The zeolite wheel's built-in PLC controller collects the actual wheel speed signal and actual desorption temperature signal in real time, and transmits the signal to the state feedback unit. The state feedback unit receives the signal and returns it to the AI algorithm controller. The AI algorithm controller optimizes the instructions again based on the feedback data, forming an energy-saving control closed loop of "device detection and analysis - zeolite wheel system's built-in PLC execution - device feedback and optimization", ensuring that the operating parameters continuously adapt to real-time working conditions.

[0023] Human-computer interaction function: The device is equipped with a human-computer interaction interface (touch screen) that can display key information: total on-time of the energy-saving control device, total energy-saving ratio of the system, total energy-saving time; real-time concentration value, condenser temperature, ambient temperature; power consumption and gas energy consumption of the zeolite wheel system per unit of time in energy-saving (device turned on) and non-energy-saving (device turned off) states; the on-off state of each module (energy-saving control device, peristaltic pump, condenser) and the zeolite wheel system; this week's daily savings cost column chart and total savings cost status bar, making it easy for maintenance personnel to monitor and manage.

[0024] Communication adaptation function: The AI algorithm controller is integrated with the zeolite wheel system's built-in PLC controller through an RS485 communication interface, or communicates remotely through an industrial Ethernet, adapting to the installation and control needs of different industrial scenarios without the need for large-scale modification of the original zeolite wheel control system.

[0025] Furnace temperature monitoring and alarm function: The device is equipped with a furnace temperature monitoring unit that communicates with the zeolite wheel system's built-in PLC controller to obtain real-time temperature parameters of the supporting furnace (such as CO catalyst furnace, RTO combustion furnace, etc.). The AI algorithm controller extracts multi-dimensional features from the collected temperature signals, including furnace type identification and real-time temperature value. Based on the pre-set type-specific temperature threshold matching rules, it determines whether the temperature parameters of the current monitoring object (CO catalyst furnace or RTO combustion furnace) meet the corresponding safe operating interval:

[0026] If it is a CO catalyst furnace, it determines whether its temperature is within the safe threshold interval of 280℃≤temperature≤580℃;

[0027] If it is a RTO combustion furnace, it determines whether its temperature is within the safe threshold interval of 760℃≤temperature≤900℃.

[0028] When the furnace temperature is monitored to exceed the safe operation interval of the corresponding type, the AI algorithm controller immediately sends an alarm signal of temperature abnormality to the PLC controller of the zeolite rotary system, and the PLC controller executes a safety protection action (such as reducing the desorption load, suspending heating, etc.) according to the preset safety protection rules of the zeolite rotary system to ensure the safe operation of the system.

[0029] Furnace temperature control function: The equipment is additionally provided with a furnace temperature control unit, which aims to solve the technical problems of "static adjustment lagging behind working condition fluctuations" and "difficulty in balancing energy consumption and processing efficiency" in the traditional zeolite rotary system furnace temperature control. The furnace temperature control unit forms a closed loop interaction with the AI algorithm controller and the PLC controller of the zeolite rotary system through signal connection, and the core function is to realize dynamic and accurate adjustment of the furnace temperature through multi-algorithm collaborative optimization - specifically through the following four layers of logic:

[0030] First layer: basic temperature prediction (BP neural network algorithm support)

[0031] The furnace temperature control unit first calls the BP neural network algorithm, and uses historical operation data (including desorption temperature, corresponding furnace temperature, associated energy consumption and waste gas treatment efficiency) as training samples to construct a mapping model of desorption temperature and furnace temperature. Based on the "historical optimal matching rule", the model outputs the basic furnace temperature parameters suitable for the current desorption demand, providing a reference for subsequent adjustment and ensuring the initial rationality of temperature control.

[0032] Second layer: dynamic correction compensation (reinforcement learning algorithm support)

[0033] For real-time working condition fluctuations (such as sudden changes in waste gas concentration and flow), the module introduces a reinforcement learning algorithm for dynamic optimization: taking the desorption multiple (the ratio of desorption air volume to treatment air volume) as the decision variable, setting the "energy consumption reduction rate" and "treatment efficiency compliance rate" as the dual reward function, and through continuous interaction learning with real-time working conditions (such as increasing the desorption multiple to strengthen the treatment effect when the concentration increases, and reducing the desorption multiple to reduce energy consumption when the concentration decreases), the optimal desorption multiple is dynamically output, and the correction range of the basic furnace temperature is calculated accordingly, realizing the precise matching of "real-time working condition-temperature adjustment".

[0034] Third layer: forward-looking collaborative optimization (MPC prediction model support)

[0035] To avoid the hysteresis problem caused by "adjustment based on current state only", the module realizes forward-looking optimization through MPC (Model Predictive Control) prediction model: integrate the basic temperature parameters output by BP neural network and the correction amplitude output by reinforcement learning, while accessing the future 5-10min exhaust condition prediction data (including concentration, flow trend), through multivariate constraint calculation (such as minimizing energy consumption under the premise of ensuring processing efficiency ≥ 90%), finally generate the most suitable furnace temperature adjustment signal that adapts to real-time working conditions and can respond to future changes.

[0036] The fourth layer: prediction data source (time series prediction model support)

[0037] The "future 5-10min exhaust condition prediction" required by the MPC model is provided by the time series prediction model. The time series prediction model integrates multiple input signals: real-time concentration signals from the concentration detection unit, real-time flow signals from the flow meter, future production scheduling (such as batch switching, load adjustment plan) provided by the target customer production planning system, and past working condition rules (such as periodic fluctuation characteristics) stored in the historical database; through time series feature extraction, trend fitting and rule matching, accurately predict the future exhaust condition changes in a short time, and transmit the prediction results to the MPC model in real time to provide data basis for forward-looking optimization.

[0038] Finally, the furnace temperature adjustment signal generated by the furnace temperature control unit is transmitted to the PLC controller built-in zeolite rotary system, and the PLC drives the furnace heating device to act, forming a complete control chain of "multi-algorithm collaborative optimization-real-time signal interaction-dynamic execution adjustment", realizing efficient, energy-saving and accurate control of the furnace temperature under various working conditions.

[0039] (3) Beneficial effects

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] Dynamic adaptation to working conditions, significant energy saving: through real-time analysis of exhaust gas concentration, flow and zeolite rotary system running state by AI algorithm, accurate control instructions are generated and connected to the PLC built-in zeolite rotary system, the rotary speed and desorption temperature are dynamically adjusted, avoiding energy consumption redundancy caused by traditional fixed parameters; under the premise of ensuring processing efficiency ≥ 90%, the total energy consumption of the zeolite rotary system can be reduced by 15%-30%, without the need for additional integration of PLC, reducing equipment development and manufacturing costs.

[0042] Accurate detection, reliable control: the gas pretreatment unit reduces the interference of impurities and humidity on detection, the PID concentration meter and the rotor flow meter ensure the detection accuracy (error ≤ 5%); combined with the multi-parameter collaborative optimization of BP neural network algorithm, the response time of the adjustment instruction is ≤ 10s, which can quickly adapt to working condition fluctuations (such as sudden increase / decrease of VOCs concentration), avoiding emission exceeding the standard or energy waste caused by instruction lag.

[0043] Closed-loop control, balance efficiency and energy saving: Through the feedback loop of "the device-zeolite wheel system with PLC", the control instruction is corrected in real time to avoid the contradiction between "low parameters leading to emission exceeding the standard" and "high parameters leading to energy waste", ensuring that the outlet VOCs concentration meets the standard (≤30 mg / m 3 ), while adapting to the original control architecture of the zeolite wheel system, with strong compatibility. Intelligent management, reducing operation and maintenance costs: The man-machine interface intuitively displays energy consumption data and equipment status; dynamic adjustment reduces the mechanical wear of the wheel (flattens the speed fluctuation) and the aging rate of the zeolite (avoids continuous high temperature), prolongs the service life of the equipment by 10%-20%, and reduces the cost of maintenance and adsorbent replacement; and no additional maintenance of PLC is required, reducing the operation and maintenance steps.

[0044] Safety improvement, strong adaptability: New furnace temperature monitoring and alarm function, timely warning of abnormal working conditions; through the standardized communication interface to connect the PLC controller of the zeolite wheel system, without modifying the original control system of the zeolite wheel, it is suitable for different brands and models of zeolite wheel systems.

[0045] Dynamic intelligent adjustment of furnace temperature solves the hysteresis of parameter adjustment: In view of the core pain point of traditional zeolite wheel systems, "the hysteresis of furnace temperature parameter adjustment to exhaust gas working condition fluctuation" - relying on manual preset fixed value or simple feedback adjustment, when facing concentration / flow mutation, often due to the lack of timely adjustment, leading to fluctuation of processing efficiency (such as below 90% of the standard threshold) or surge of energy consumption (such as excessive heating), the invention adds a furnace temperature control unit to build an intelligent adjustment system of "multi-algorithm cooperation + forward-looking prediction + closed-loop control", realizing the upgrade from traditional "extensive, passive, empirical" control to "precise, proactive, intelligent" control.

[0046] Specifically, this system breaks through the hysteresis bottleneck through three layers of technical innovation: first, use BP neural network to mine historical data rules to output basic temperature parameters that adapt to current desorption needs, laying the foundation for adjustment; second, use reinforcement learning to respond to working condition fluctuations (such as concentration jump, flow change) in real time, dynamically adjusting the temperature adjustment range to achieve "working condition changes, adjustment changes" in real time; third, rely on the MPC model to integrate the previous results, and combine the future 5-10 minutes of exhaust gas working condition prediction (concentration, flow trend) provided by the time series prediction model to layout temperature adjustment strategies in advance, making the control action "run ahead of working condition changes". At the same time, the module, AI algorithm controller, and PLC form a signal closed loop to ensure that there is no delay in the whole process from calculation to execution of the adjustment instruction, ultimately realizing three core effects:

[0047] More stable processing efficiency: Compared with traditional control methods, the fluctuation range of processing efficiency is reduced by more than 30%, and it is long-term stable in the standard interval of more than 90%;

[0048] Lower energy cost: By precisely matching temperature and working condition requirements, furnace energy consumption (such as gas / electricity) is reduced by 15%-30%;

[0049] Stronger adaptability and automation level: No need for manual intervention to deal with complex working conditions such as production plan adjustment and periodic fluctuations, completely free from dependence on operating experience.

[0050] This technical solution is particularly suitable for scenarios where industrial waste gas composition is complex, working conditions fluctuate frequently, and both energy saving and emission standard compliance are strictly required (such as chemical, coating, printing, automobile, shipbuilding, etc.). BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is the overall architecture diagram of the present application;

[0052] Figure 2 is the closed-loop control flowchart of the present application;

[0053] Figure 3 is the control unit AI algorithm controller logic flowchart of the present application;

[0054] Figure 4 is the communication connection diagram of the present application;

[0055] Figure 5 is the furnace temperature detection and alarm logic flowchart of the present application;

[0056] Figure 6 is the overall connection architecture diagram of the furnace temperature control unit and system of the present application;

[0057] Figure 7 is the algorithm coordination logic flowchart of the furnace temperature control unit of the present application;

[0058] Figure 8 is the time series prediction model data input and output logic flowchart of the present application.

[0059] Figure 9 is the man-machine interface diagram of the present application

[0060] Figure 10 is the man-machine interface display communication architecture diagram of the present application DETAILED DESCRIPTION

[0061] The technical solutions of the present application will be described below in conjunction with examples.

[0062] Example 1, combined Figure 1The system architecture diagram shown, the energy-saving control device (100) of the application realizes the energy-saving optimization and furnace safety control of the industrial waste gas treatment process through the technical logic of "multi-unit cooperative collection + AI algorithm intelligent decision + linkage with zeolite rotary system (original equipment, not a component of the application)", and the specific implementation is as follows:

[0063] The pre-filtering unit (110) extracts a small amount of industrial waste gas at a flow rate of 1.5-3 L / min through the suction pump (112), and completes the pretreatment according to the "two-stage filtration + condensation and dehumidification" process to meet the accuracy requirements of subsequent concentration detection:

[0064] The waste gas is first preliminarily filtered by the PE material polymer filter 1 (111, filtering accuracy 1-10 μm) to intercept large-particle impurities (such as dust and solid pollutant residues) in the waste gas with a particle size of ≥1 μm; the intercepted impurities contain a small amount of sticky components, which are extracted and discharged by the peristaltic pump (115) at a pumping and discharging flow range of 0.1-3 L / h in a timely manner to avoid clogging of the filter 1 internal filter material by impurity accumulation;

[0065] The waste gas is continuously conveyed by the suction pump (112) to the PE material polymer filter 2 (113, filtering accuracy 1-10 μm) for secondary filtration to further remove residual small-particle impurities (non-sticky solid impurities); the filter 2 is provided with a blowdown port at the bottom, and the impurities are discharged to the system blowdown pipeline by gravity flow without the need for additional power;

[0066] The waste gas after two-stage filtration enters the condenser (114), which stabilizes the gas temperature at 5-10°C through temperature closed-loop control, and simultaneously reduces the relative humidity of the waste gas to ≤30% (to avoid the interference of high humidity environment on the detection accuracy of the PID concentration instrument), and the gaseous water in the waste gas is condensed into liquid sewage by the condenser; the condenser is provided with a drainage structure (such as a drain valve) inside, which uses the slight positive pressure formed inside during operation (accompanied by slight air pressure during the desorption heating process) to discharge the sewage to the blowdown pipeline, avoiding liquid accumulation;

[0067] The discharges of the peristaltic pump (115), the filter 2 blowdown port and the condenser blowdown port are finally collected into the same set of blowdown pipeline for unified treatment. The clean waste gas after pretreatment is conveyed to the concentration detection unit (120) through a VOCs corrosion-resistant hose.

[0068] The core components (filter 1, suction pump, filter 2, condenser, peristaltic pump, flow meter) of the pre-filtering unit (110) and the concentration detection unit (120) are connected by a VOCs corrosion-resistant hose (inner diameter 10-15 mm) through a sealed connection mode of double clamp sealing. Specifically, the two ends of the hose are fixed by 316L stainless steel double clamps, lined with FKM fluororubber sealing rings (temperature resistance -20~200 DEG C, benzene series swelling rate ≤5%), ensuring no leakage under the working conditions of -10 DEG C to 50 DEG C. The hose material is suitable for the corrosion environment of industrial VOCs waste gas, avoiding pipeline leakage after long-term use; the hose connection adopts a tee joint to realize turning, and the local gas flow is adjusted by a flow limiting valve to ensure stable gas flow between units.

[0069] II. Signal acquisition of the concentration detection unit (reference numeral 120)

[0070] The concentration detection unit (120) obtains key working condition parameters of waste gas through the combination of "flow control + concentration detection", providing real-time data support for AI decision-making:

[0071] The pretreated waste gas first flows through the flow meter (121, rotor flow meter (range 0.5-5 m 3 / h, flow control accuracy ±2.5% FS), detects and outputs the real-time flow signal of the waste gas (ensures stable gas flow into the concentration meter, avoiding flow fluctuation leading to concentration detection deviation);

[0072] The waste gas with stable flow enters the PID concentration meter (122, detection accuracy ±5% FS, response time ≤10 s, with zero point automatic calibration function), which detects the concentration signal of the target pollutants (such as non-methane total hydrocarbons, benzene series, etc.) in the waste gas in real time;

[0073] The concentration detection unit (120) synchronously transmits the "real-time flow signal + pollutant concentration signal" to the control unit (150) through the internal communication line.

[0074] III. Signal acquisition of the state feedback unit (130)

[0075] The state feedback unit (130) is the "signal interaction interface between the present application and the zeolite rotary system", which receives two types of core operating signals transmitted by the PLC controller (F140) of the zeolite rotary system (original, not all of the present application) (F100):

[0076] Zeolite runner rotation speed signal: collected by the original runner rotation speed sensor (F131) of the zeolite runner system (connected with the runner driving motor (F121) of the actuator (F120)), processed by the PLC controller (F140), transmitted to the "zeolite runner rotation speed signal module (131)" of the state feedback unit (130), and then transmitted to the control unit (150) after standardization processing by the module.

[0077] Zeolite runner desorption temperature signal: collected by the original desorption temperature sensor (F132) of the zeolite runner system (installed in the desorption air duct and connected with the desorption heating device (F122) of the actuator (F120)), converted by the PLC controller (F140), transmitted to the "desorption temperature signal module (132)" of the state feedback unit (130), and then transmitted to the control unit (150) after standardization.

[0078] Four, AI algorithm decision and instruction output of the control unit (label 150)

[0079] The AI algorithm controller (151) of the control unit (150) is the core decision module, which is built-in machine learning algorithm, with "waste gas treatment efficiency ≥ 90% as the premise, minimizing the comprehensive energy consumption of the equipment" as the optimization goal, and the specific logic is as follows:

[0080] Algorithm training basis: using historical operation data as training samples, the samples include "historical waste gas concentration, flow, runner rotation speed, desorption temperature" and "electricity consumption, gas consumption" in the corresponding period, to ensure that the algorithm has "working condition-energy consumption" correlation learning ability;

[0081] Real-time signal input: the AI algorithm controller (151) synchronously receives three types of signals:

[0082] "Real-time flow signal + pollutant concentration signal" of the concentration detection unit (120);

[0083] "Standardized runner rotation speed signal + standardized desorption temperature signal" of the state feedback unit (130);

[0084] Instruction generation and transmission: through real-time algorithm calculation, "runner rotation speed adjustment instruction" and "desorption temperature adjustment instruction" (i.e. optimal parameter instruction) are generated to adapt to the current working condition, and transmitted to the PLC controller (F140) of the zeolite runner system through the communication module (160) —— among them, the communication module (160) is built-in switch, supporting RS485 interface, industrial Ethernet (Profinet protocol) interface, communication rate ≥ 9600bps, to ensure the stability and real-time performance of the instruction transmission.

[0085] Five, action control of the zeolite runner system actuator

[0086] The PLC controller (F140) of the zeolite rotary system serves as a "command execution relay node" and forms a signal closed loop with the AI algorithm controller (151) and the zeolite rotary system execution mechanism (F120):

[0087] After receiving the "rotary speed adjustment command" and the "desorption temperature adjustment command", the PLC controller (F140) converts them into electrical signals (such as analog 4-20mA and digital switch signals) suitable for the execution mechanism (F120);

[0088] The two types of execution mechanisms are driven respectively:

[0089] The rotary speed adjustment signal is sent to the "rotary drive motor (F121)", changing the rotation rate of the zeolite rotary (F111) (adapted to the change of exhaust gas concentration, avoiding adsorption saturation or excessive energy consumption);

[0090] The power adjustment signal is sent to the "desorption heating device (F122)", adjusting the heating temperature of the desorption air duct (ensuring VOCs resolution efficiency, while avoiding overheating and wasting energy);

[0091] Finally, the "on-demand adjustment" of the exhaust gas treatment process is realized, achieving the energy-saving optimization control goal of "efficiency standard + lowest energy consumption".

[0092] Six, collection of furnace temperature, safety alarm and dynamic adjustment

[0093] 1. Furnace temperature collection

[0094] The furnace temperature monitoring unit (140) collects the temperature signal of the zeolite rotary system supporting furnace (F112), which refers to the CO catalytic furnace (represented by CO in the figure) or the RTO combustion furnace (represented by RTO in the figure):

[0095] The furnace temperature is collected by the original furnace temperature sensor (F133, PT100 platinum resistance temperature sensor) of the zeolite rotary system and transmitted to the PLC controller (F140) of the zeolite rotary system;

[0096] After the PLC controller (F140) converts the furnace temperature signal into a standard communication signal, it is transmitted to the furnace temperature monitoring unit (140);

[0097] After the furnace temperature monitoring unit (140) performs noise reduction processing on the signal, it is transmitted to the AI algorithm controller (151) of the control unit (150).

[0098] 2. Furnace temperature safety alarm (combined with Figure 5 )

[0099] AI algorithm controller (151) built-in sub-type temperature threshold decision rules:

[0100] If the monitoring object is a CO catalytic furnace, determine whether the temperature is in the safety interval of 280℃≤T≤580℃;

[0101] If the monitoring object is a RTO combustion furnace, determine whether the temperature is in the safety interval of 760℃≤T≤900℃;

[0102] If the temperature exceeds the safety interval (exceeds the upper limit or the lower limit), the AI algorithm controller (151) immediately generates a "temperature abnormal alarm signal" (including the type of abnormality and the real-time temperature value) and feeds back to the PLC controller (F140) of the zeolite rotary system, which executes the preset safety protection action (such as shutdown, load reduction, fuel supply cutoff, etc., the specific logic is shown in the subsequent embodiments). Figure 6

[0103] 3. Furnace temperature dynamic adjustment (combined with Figure 6 , Figure 7 , Figure 8 )

[0104] The AI algorithm controller (151) generates an adjustment strategy based on the real-time temperature signal of the furnace temperature monitoring unit (140) through multi-algorithm collaborative optimization:

[0105] The collaborative algorithms include BP neural network algorithm (predicting the basic furnace temperature), reinforcement learning algorithm (dynamically calculating the temperature correction amplitude), MPC (model predictive control) algorithm (integrating multi-parameter optimization), time series prediction model (predicting future working conditions), and other industrial furnace temperature optimization algorithms known in the art (such as PID parameter self-tuning algorithm);

[0106] Finally, a "furnace temperature adjustment signal" is generated and transmitted to the PLC controller (F140) of the zeolite rotary system through the communication module (160), which controls the furnace heating device (F123, such as a gas proportional valve) to act, realizing the dynamic and accurate regulation and control of the furnace temperature (the specific logic is shown in the subsequent embodiments). Figure 7 , Figure 8 , Figure 9

[0107] Seven, human-computer interaction interface:

[0108] In the control unit (150), a human-computer interaction interface (reference numeral 152) is also integrated; the human-computer interaction interface is a touch screen, which is connected with the AI algorithm controller (151) through an internal bus (such as RS485 bus) to realize signal connection, and is used for visual display of system running state and energy saving data, and the specific display content includes but is not limited to:

[0109] ​​Energy consumption information: the amount of electric energy consumption (unit: kW·h) and the amount of gas consumption (unit: m 3 ) of the zeolite rotary device in unit time (such as per hour) under the conditions of "the device is turned on (energy-saving state)" and "the device is turned off (non-energy-saving state)";

[0110] Device state information: the energy-saving on state of the zeolite rotary device, the on state of the energy-saving control device, the on state of the peristaltic pump, and the on state of the condenser, which include two modes of real-time running / off state. The running / off state is presented in the form of "running" "off" text or indicator icon;

[0111] Statistical information: total on time of the zeolite rotary device, total energy-saving ratio of the system, total energy-saving time, column chart of daily savings of the current week, and status bar of total savings of the current week;

[0112] Real-time parameter information: real-time concentration value of waste gas, working temperature value of the condenser, and environmental temperature value. The specific embodiments of the human-computer interaction interface (152) will be described in detail in Figure 9 、 Figure 10 Embodiment 9.

[0113] In addition, it should be noted that the state feedback unit (130) and the furnace temperature monitoring unit (140) are built-in functional modules (without independent physical hardware entities, relying on the software logic of the AI algorithm controller) of the AI algorithm controller (151): by establishing communication with the PLC controller (F140) of the zeolite rotary system, the signal reception and standardization processing are completed, and data support is provided for the decision and optimization of the AI algorithm, without the need for additional configuration of independent hardware interface.

[0114] Embodiment 2, in combination with the closed-loop control flowchart shown in Figure 2 , the specific implementation process is as follows:

[0115] 1. Collection and feedback of real-time running parameters (step 210)

[0116] The state feedback unit receives the real-time running parameters returned by the PLC controller of the zeolite rotary system, including the rotation speed and desorption temperature of the zeolite rotary device. These parameters are the core representation of the current running state of the zeolite rotary system, and provide basic data input for subsequent algorithm calculation.

[0117] 2. Optimization calculation of the AI algorithm controller (step 220)

[0118] The AI algorithm controller receives the real-time operating parameters transmitted by the state feedback unit and performs calculation based on the built-in BP neural network algorithm model. The model takes "waste gas treatment efficiency meeting the standard and equipment energy consumption minimization" as the optimization goal, and through the training of historical operating data (such as waste gas concentration, flow and corresponding energy consumption under different rotating speeds and desorption temperatures), it can dynamically generate the optimal control strategy according to the real-time working conditions. After receiving the current rotating speed, desorption temperature and other parameters, the model quickly operates to generate the optimization instructions (including specific parameters for rotating speed adjustment and temperature adjustment) for the zeolite rotating wheel system.

[0119] 3. Transmission of control instructions (step 230)

[0120] The communication module receives the optimization instructions generated by the AI algorithm controller and transmits them to the PLC controller built-in the zeolite rotating wheel system. The communication module ensures the real-time and reliability of the instruction transmission through the adapted communication protocol (such as industrial Ethernet, RS485, etc.), and provides a bridge for the PLC controller to issue and execute instructions.

[0121] 4. Action execution of the actuator (step 240)

[0122] After receiving the instructions from the communication module, the PLC controller built-in the zeolite rotating wheel system issues control instructions to the actuators:

[0123] For the rotating wheel driving motor, the instruction adjusts the rotating speed to adapt to the current waste gas treatment requirements for the rotating wheel adsorption / desorption rhythm;

[0124] For the desorption heating device, the instruction adjusts the temperature to ensure efficient and reasonable energy consumption in the desorption process.

[0125] 5. Closed-loop iteration of system state (step 250)

[0126] After the actuators act according to the instructions, the operating state of the zeolite rotating wheel system is directly changed (such as rotating speed change, desorption temperature adjustment). At this time, the new real-time operating parameters of the zeolite rotating wheel system (updated rotating speed, desorption temperature, etc.) will be fed back to the state feedback unit again, entering the next "collection-calculation-control-feedback" cycle, so as to realize the continuous optimization of the operating efficiency and energy consumption of the zeolite rotating wheel system.

[0127] Example 3: BP neural network parameter optimization process of the control unit AI algorithm controller (combined with Figure 3 )

[0128] combined with Figure 3The control unit AI algorithm controller logic diagram is shown, and the structure, training process and signal flow logic of the BP neural network layer (320) are described in detail in this embodiment, so that those skilled in the art can reproduce the technical solution of “multi-parameter input → AI optimization calculation → speed / temperature regulation instruction output”.

[0129] I. Definition of BP neural network hierarchy and input and output

[0130] The BP neural network layer (320) adopts a full connection architecture of “input layer (321) → hidden layer 1 (322) → hidden layer 2 (323) → output layer (324)”, and the number of neurons, input and output parameters and processing logic of each layer are as follows:

[0131] Input layer (321): contains 4 neurons, receives input signal (310), specifically:

[0132] Concentration signal (311): real-time concentration of VOCs (unit: mg / m 3 );

[0133] Flow signal (312): real-time flow of exhaust gas (unit: m 3 / h);

[0134] Actual rotation speed of zeolite runner (313): current running speed of runner (unit: r / min);

[0135] Actual desorption temperature (314): current temperature of desorption device (unit: ℃).

[0136] The input signal needs to be standardized (normalized to the [0, 1] interval by the formula x norm = (x-μ) / σ, where μ is the mean of the training set and σ is the standard deviation of the training set), and then input into the neuron.

[0137] Hidden layer 1 (322) and hidden layer 2 (323):

[0138] Hidden layer 1 (322): contains 12 neurons, adopts Sigmoid activation function (formula: f(x) = 1 / (1+e^(-x))), performs nonlinear mapping on the linear combination signal transmitted by the input layer, and captures the complex correlation between parameters;

[0139] Hidden layer 2 (323): contains 12 neurons, also adopts Sigmoid activation function, further performs nonlinear transformation on the output of hidden layer 1, and enhances the fitting ability of the model to the “concentration-flow-speed-temperature” multi-parameter coupling relationship.

[0140] The signal is transmitted between neurons in each layer through a weight matrix (the initial value of the weight matrix uses the Xavier initialization method to avoid gradient disappearance or explosion during the training process).

[0141] Output layer (324): contains 2 neurons, outputs output instruction (330), specifically:

[0142] Turbine speed adjustment value (331): target speed adjustment amount of zeolite turbine (unit: r / min);

[0143] Desorption temperature adjustment value (332): target temperature adjustment amount of desorption heating device (unit: °C).

[0144] The output value needs to be processed by inverse standardization (reduced to the actual physical quantity range through the formula x raw = x norm ×σ+μ), and then output as a control instruction.

[0145] II. Training process of BP neural network

[0146] In order to make the network learn the correlation between "concentration, flow, current speed / temperature and optimal speed adjustment, temperature adjustment", the network weight needs to be trained and optimized through historical data. The specific process is as follows:

[0147] Training data set construction: collect historical operation data of the system in the past 12 months, select 50,000+ valid samples, covering the full operating condition range:

[0148] VOCs concentration: 50-1500mg / m 3 ;

[0149] Exhaust gas flow: 2000-12000m 3 / h;

[0150] Among them, the abnormal working condition sample ratio of concentration mutation (±30% / 10min) is 20%, in order to enhance the adaptability of the algorithm to fluctuating conditions.

[0151] Each sample contains "input layer 4 parameters (311-314)" and "manually labeled optimal output layer parameters (331-332, determined based on expert experience or historical optimal energy efficiency operating conditions)".

[0152] Data preprocessing:

[0153] Missing value processing: linear interpolation method is used to fill in the missing values in time series data;

[0154] Outlier processing: for abnormal values exceeding "mean value ± 3 times standard deviation", truncate to a reasonable range of operating conditions (such as setting the upper limit of concentration to 1500mg / m 3, the lower limit is 50 mg / m 3 );

[0155] Standardization: Standardize the input and output parameters according to "x norm = (x - μ) / σ", where μ and σ are based on the statistics of the training set.

[0156] Specifically: x in the molecule: represents the original input data (such as the concentration of VOCs collected at a certain time, the flow of waste gas, etc. The original numerical value is not processed); x norm represents the standardized data (the original data x

[0157] After the normalization processing of "subtracting the mean and dividing by the standard deviation", the numerical value of the adaptive neural network input range is obtained); μ: the average value of the corresponding feature (such as the "concentration" feature) in the training data set;

[0158] σ: the standard deviation of the corresponding feature in the training data set (measures the dispersion of the feature data).

[0159] Training parameters and optimization logic:

[0160] Loss function: Mean Squared Error (MSE) is used, the formula is

[0161] MSE = 1 / n∑ n i＝1 (y1i-y2i)2(y1i is the labeled optimal output, y2i is the network prediction output, n is the sample number);

[0162] Optimization algorithm: Adam optimizer is used, the initial learning rate is set to 0.001, and the learning rate is attenuated to 0.9 times of the original after every 500 iterations;

[0163] Training rounds: a total of 10,000 training rounds, when the MSE changes less than 1e-5 for 200 consecutive rounds, the training is terminated in advance;

[0164] Verification mechanism: divide the data set into training set and validation set according to "8:2", if the average absolute error (MAE) of the validation set is > 3 (the MAE of the speed is > 3 r / min or the MAE of the temperature is > 3 ℃), adjust the number of hidden layer neurons (such as increase to 16) and retrain until the MAE of the validation set is ≤ 3.

[0165] Three, signal flow and intelligent optimization implementation

[0166] The flow and calculation process of the input signal (310) is as follows:

[0167] The concentration signal (311), the flow signal (312), the actual rotation speed of the zeolite runner (313), and the actual desorption temperature (314) are input into the input layer (321) of the BP neural network after standardization processing.

[0168] The four neurons of the input layer (321) transmit signals to the hidden layer 1 (322) and perform nonlinear transformation through a Sigmoid activation function.

[0169] The output of the hidden layer 1 (322) is transmitted to the hidden layer 2 (323) and is transformed again through a Sigmoid activation function.

[0170] The output of the hidden layer 2 (323) is transmitted to the output layer (324) and is output as the runner speed adjustment value (331) and the desorption temperature adjustment value (332) after inverse standardization processing.

[0171] The output instruction (330) is directly used to guide the speed adjustment of the zeolite runner and the temperature adjustment of the desorption heating device, realizing AI intelligent optimization under multiple parameters.

[0172] Embodiment 4, in combination Figure 4 As shown in the communication connection schematic diagram, the energy-saving control device (100) of the application realizes intelligent collaborative control with the original zeolite runner system through a bidirectional communication closed loop, and the specific implementation process is as follows:

[0173] 1. Overall architecture logic

[0174] The AI algorithm controller (151) and the communication module (160) in the energy-saving control device (100) are integrated through an internal bus to complete data interaction; the communication module (160) provides two communication schemes, both of which support bidirectional signal transmission with the PLC controller (F140) of the original zeolite runner system (F100), forming a closed loop control logic of "PLC operation parameter feedback AI algorithm calculation AI instruction issuance PLC execution", which provides a basis for the "real-time state optimization control" of the AI algorithm.

[0175] 2. Scheme one: RS485 local connection (short distance, low delay scene)

[0176] The AI algorithm controller (151) is connected with the RS485 communication interface (AI controller side, 161) in the communication module (160) through internal bus integration; then, it is bidirectionally interconnected with the RS485 communication interface (PLC side, 162) through a shielded twisted pair, and finally it is connected to the PLC controller of the original zeolite runner system.

[0177] The signal interaction logic is as follows:

[0178] Signal 1 (PLC→AI): The PLC controller of the zeolite rotary system transmits real-time operation parameters (such as actual rotary speed of the zeolite rotary, real-time temperature of the desorption zone, etc.) to the AI algorithm controller (151) through the RS485 communication interface (162), shielded twisted pair, and RS485 communication interface (161), providing "real-time state input" for the energy-saving optimization calculation of AI.

[0179] Signal 2 (AI→PLC): The AI algorithm controller (151) obtains optimal adjustment instructions (such as target value of rotary speed of the rotary, target value of desorption temperature) based on the transmitted operation parameters, and then transmits the instructions to the PLC controller through the above-mentioned RS485 link, driving the actuators (such as rotary drive motor, desorption heating device) of the zeolite rotary system to act.

[0180] 3. Scheme two: Industrial Ethernet remote connection (long-distance, flexible networking scene)

[0181] The AI algorithm controller (151) is connected with the industrial Ethernet interface (AI controller side, 163) in the communication module (160) through an internal bus, and then is bidirectionally interconnected with the industrial switch (164) through a CAT6 / 7 network cable, and the industrial switch (164) is bidirectionally interconnected with the industrial Ethernet interface (PLC side, 165) through a CAT6 / 7 network cable, and finally is connected to the PLC controller (F140) of the zeolite rotary system. The communication module (160) is built-in switch, which is used for realizing signal collection and stable transmission between the AI algorithm controller and the PLC controller (F140) of the zeolite rotary system, supporting RS485 interface and industrial Ethernet (Profinet protocol) communication, and the communication rate is greater than or equal to 9600 bps.

[0182] The signal interaction logic is consistent with that of scheme one:

[0183] Signal 1 (PLC→AI): The PLC controller (F140) transmits real-time operation parameters to the AI algorithm controller (151) through the industrial Ethernet interface (165), the industrial switch (164), and the industrial Ethernet interface (163);

[0184] Signal 2 (AI→PLC): The AI algorithm controller (151) transmits optimal adjustment instructions to the PLC controller (F140) through the above-mentioned industrial Ethernet link, and completes intelligent optimization of the operation parameters of the zeolite rotary system.

[0185] Two communication schemes can be flexibly adapted to different on-site networking requirements (such as selecting RS485 for a short-distance scene to reduce costs, and selecting industrial Ethernet for a long-distance / multi-device scene to improve networking flexibility); at the same time, with the help of bidirectional signal interaction, it is ensured that the AI algorithm can continuously optimize the control strategy based on the real running data of the system, which not only guarantees the stable operation of the zeolite rotary system, but also maximizes the energy saving efficiency.

[0186] Embodiment 5, as shown in the furnace temperature detection and alarm logic diagram, the energy-saving control device of the present application realizes the temperature monitoring and alarm function of the furnace matched with the zeolite rotary system (original) through the following process, and the specific implementation is as follows: Figure 5

[0187] 1. Start (step 510)

[0188] The energy-saving control device (the present application) receives a start instruction (such as a field trigger or a remote control signal), and formally starts the process of furnace temperature monitoring and AI control.

[0189] 2. Real-time acquisition of temperature signal (step 520)

[0190] The furnace temperature monitoring unit configured by the device establishes bidirectional communication with the PLC controller of the zeolite rotary system; the PLC controller of the zeolite rotary system continuously transmits the real-time temperature signal collected by the PT100 platinum resistance temperature sensor of the furnace (CO catalytic furnace or RTO combustion furnace) to the furnace temperature monitoring unit, and completes the real-time acquisition of temperature data.

[0191] 3. Temperature signal feature extraction (step 530)

[0192] After the AI algorithm controller receives the temperature signal transmitted by the furnace temperature monitoring unit, it performs multi-dimensional feature extraction on it, and the extraction content includes:

[0193] Furnace type identification (distinguish whether the current monitored object is a CO catalytic furnace or a RTO combustion furnace);

[0194] Real-time temperature value of the furnace.

[0195] 4. Type-specific temperature threshold determination (step 540)

[0196] The AI algorithm controller determines whether the temperature parameter of the currently monitored furnace meets the corresponding safety threshold interval based on the preset type-specific temperature threshold matching rule.

[0197] 5. Determination standard of type-specific temperature threshold matching rule (550)

[0198] If the monitored object is a CO catalytic furnace, it is determined whether its temperature is in the safety threshold interval of 280℃≤temperature≤580℃;​

[0199] If the monitoring object is the RTO combustion furnace, it is determined whether its temperature is in the safety threshold interval of 760℃≤temperature≤900℃. 4. Normal operation branch: algorithm parameter optimization (steps 560, 570)

[0200] When it is determined that the furnace temperature is in the corresponding safety threshold interval:

[0201] The AI algorithm controller receives the “normal temperature data (real-time temperature value)” and uses it for parameter optimization of the subsequent furnace temperature control algorithm.

[0202] The zeolite wheel system (original) and the energy-saving control device of the present application continue to operate normally according to the normal logic, without triggering the safety protection action.

[0203] 5. Abnormal operation branch: transmission of alarm signal (steps 580, 590 and F510, F520, F530 executed by the zeolite wheel system (original) (F100))

[0204] When it is determined that the furnace temperature is out of the corresponding safety threshold interval:

[0205] Step 580: The AI algorithm controller triggers the alarm logic to generate an abnormal alarm signal containing the temperature abnormality type (over the upper limit or over the lower limit) and the real-time temperature value.

[0206] Step 590: The abnormal alarm signal is transmitted to the communication module and forwarded to the PLC controller of the zeolite wheel system by the communication module.

[0207] Step F510: The PLC controller receives the alarm signal

[0208] Step F520: After receiving the alarm signal, the PLC controller executes the corresponding safety strategy (such as reducing the desorption load, suspending heating, etc.) according to the safety protection rules preset by the zeolite wheel system.

[0209] Step F530: The system enters the alarm state and waits for manual intervention to ensure the safety of the equipment.

[0210] Example 6, in combination Figure 6 The energy-saving control device (100) of the present application integrates multiple types of algorithm optimization logic through the AI algorithm controller (151) to realize dynamic adjustment of the furnace temperature, and the specific implementation is as follows:

[0211] 1. Multi-source signal acquisition and input

[0212] The three types of units built-in the energy-saving control device (100) synchronously transmit real-time signals to the AI algorithm controller (151):

[0213] Concentration detection unit (120): Collects and transmits exhaust gas concentration signals and exhaust gas flow signals;

[0214] Status feedback unit (130): collects and transmits zeolite rotor speed signal and desorption temperature signal;

[0215] Furnace temperature monitoring unit (140): Collects and transmits furnace temperature signals.

[0216] 2. AI Algorithm Collaborative Optimization and Regulation Signal Generation

[0217] The AI ​​algorithm controller (151) performs fusion analysis and dynamic optimization on the input multi-source signals based on preset multi-type collaborative algorithms (such as BP neural network, reinforcement learning, MPC model predictive control, etc.), calculates the furnace temperature control strategy adapted to the real-time operating conditions, and generates the furnace temperature adjustment signal.

[0218] 3. Command transmission and furnace temperature regulation execution

[0219] The furnace temperature regulation signal is transmitted to the PLC controller (F140) built into the zeolite rotor system (existing) (F100); after receiving the regulation signal, the PLC controller (F140) generates the corresponding furnace temperature regulation command (such as the gas proportional valve opening control command) and sends it to the furnace heating device (F123 such as the gas proportional valve).

[0220] After receiving the command, the furnace heating device (F123) performs adjustment actions (such as adjusting the opening of the gas proportional valve), and finally achieves dynamic and precise control of the temperature of the matching furnace (F112, CO catalytic furnace (represented by CO in the figure) or RTO combustion furnace (represented by RTO in the figure)).

[0221] Example 7, as Figure 7 The diagram illustrates the process of AI algorithm controller collaboratively optimizing furnace temperature. The energy-saving control device of this invention achieves dynamic and precise adjustment of furnace temperature through the logic of "multi-algorithm parallel processing + multi-source data fusion + collaboration with external systems." The specific implementation method is as follows:

[0222] I. Multi-source data input: providing the algorithm with "historical benchmarks" and "real-time feedback"

[0223] The “historical operating data” module labeled 701 inputs historical desorption temperature, furnace temperature, and energy consumption data and exhaust gas treatment efficiency data associated with the temperature into the collaborative optimization computing core (703) of the AI ​​algorithm controller, providing the algorithm with a benchmark training sample of “historical operating conditions-temperature-energy efficiency”.

[0224] The "real-time working condition data" module of label 702 synchronously inputs real-time parameters such as exhaust gas concentration fluctuation and flow rate change into the collaborative optimization calculation core (703), to provide dynamic feedback of "current working condition" for the algorithm.

[0225] II. "Dual-algorithm parallel processing" of the AI algorithm core: The collaborative optimization calculation core (703) of the AI algorithm controller for generating temperature reference and dynamic correction amount is built-in with two types of algorithms for parallel analysis and processing of input data: BP neural network algorithm (7031):

[0226] The historical running data of the system in the past 12 months are taken as the training set, which covers the working condition range of VOCs concentration

[0227] 50-1500mg / m 3 , and exhaust gas flow rate 2000-12000m 3 / h, containing more than 50,000 samples (of which the abnormal working condition samples with concentration mutation account for 20%). After training, the algorithm establishes a mapping model of "desorption temperature and furnace temperature", learns the correlation between the basic furnace temperature and energy efficiency under different desorption requirements, and finally outputs the basic furnace temperature parameter (704) that meets the current desorption requirement.

[0228] Reinforcement learning algorithm (7032):

[0229] Taking desorption multiple (the ratio of desorption air volume to exhaust gas treatment air volume) as the decision variable, and taking "energy consumption reduction rate" and "exhaust gas treatment efficiency compliance rate" as the reward function, the algorithm interacts with "real-time working condition data (concentration fluctuation, flow rate change)" to dynamically output the optimal desorption multiple and temperature correction amplitude (705) (i.e. the fine correction amount that needs to be made to the "basic furnace temperature" to balance the "energy consumption reduction rate" and the "exhaust gas treatment efficiency compliance rate" under the current dynamic working condition).

[0230] The "MPC model predictive control" module of label 707 integrates three types of inputs for "multi-parameter fusion + constraint optimization calculation":

[0231] Input 1: the basic furnace temperature parameter (704) output by the BP neural network (temperature reference value);

[0232] Input 2: the optimal desorption multiple and temperature correction amplitude (705) output by the reinforcement learning (dynamic correction amount);

[0233] Input 3: "future 5-10 minute exhaust gas working condition prediction" provided by the external time sequence prediction module (such as the trend of exhaust gas concentration and flow rate change, to provide the basis for "advance prediction and pre-adjustment").

[0234] Through the "time dimension prediction + multi-target constraint optimization" of the above parameters, the MPC model finally generates the most suitable furnace temperature adjustment signal (708) (i.e., the "furnace temperature control instruction that not only matches the current desorption demand, but also adapts to future working condition changes, and balances energy consumption and efficiency").

[0235] Four, signal interaction with external systems: drive furnace temperature adjustment execution

[0236] The energy-saving control device of the present application transmits the "furnace temperature adjustment signal (708)" to:

[0237] The PLC controller (F140, not part of the present application) that comes with the zeolite rotary system: the PLC controller, as the control core of the external system, generates corresponding furnace heating device control instructions after receiving the signal;

[0238] Furnace heating device (F123, not part of the present application, such as gas proportional valve): after receiving the instructions from the PLC controller, it performs adjustment actions (such as adjusting the opening of the gas proportional valve), and finally realizes dynamic and accurate adjustment of the furnace temperature (F701, adjustment effect of the external system).

[0239] Supplementary notes:

[0240] The PLC controller (F140) and the furnace heating device (F123) belong to the original equipment of the zeolite rotary system and are not part of the present application; the dynamic and accurate adjustment of the furnace temperature (F701) is the temperature adjustment effect achieved by the zeolite rotary system (original) after receiving the output signal of the present application - the core contribution of the present application is to "generate accurate furnace temperature adjustment signals through multi-algorithm cooperation", and indirectly realize the optimization control of the furnace temperature through signal interaction with the external system.

[0241] Example 8, combined with Figure 8 The time series prediction model working logic flowchart shown in the figure, the energy-saving control device of the present application realizes future waste gas working condition prediction through multi-source data fusion and time series prediction algorithm, the specific implementation manner is as follows:

[0242] One, multi-source data input: provide comprehensive basis for time series prediction

[0243] The data input source (801) includes four types of signals:

[0244] Real-time concentration signal (8011) output by the concentration detection unit;

[0245] Real-time flow signal (8012) output by the flowmeter;

[0246] The production planning system provides future production scheduling (8013, determined by the target customer's own production planning), and the specific implementation is that the production planning system pushes the production scheduling of the next 24 hours in real time through an API interface, including four fields of "work order start time (accurate to minutes), process type (such as painting / printing), expected exhaust flow (m 3 / h), and expected VOCs concentration (mg / m 3 )", which are updated automatically every 30 minutes as exogenous variable inputs for the time series prediction model;

[0247] The historical database stores past working condition rules (8014).

[0248] II. Analysis and calculation of the time series prediction model

[0249] The above four types of signals are input into the time series prediction model (802, optional algorithms include ARIMA, LSTM, Prophet, etc.).

[0250] The time series prediction model (802) extracts time series features, fits trends, and matches rules based on preset time series analysis rules, and analyzes the exhaust working condition trend in the future period through algorithm operation.

[0251] III. Output of future exhaust working condition prediction results

[0252] The time series prediction model (802) finally outputs the exhaust working condition prediction (803) of the next 5-10 minutes, which includes the change trend of exhaust concentration and flow in the future period.

[0253] IV. Signal interaction with the MPC model

[0254] As shown in Figure 8 , the above "future 5-10 minute exhaust working condition prediction" (803) is transmitted to the "MPC model predictive control" module (707, corresponding to the "MPC model predictive control" module in Figure 8 ), which provides the basis for "future working condition prediction" for the subsequent "multi-parameter integration and furnace temperature regulation signal generation".

[0255] Example 9: Human-computer interaction display system (communication architecture and display logic)

[0256] Combined with Figure 9 the human-computer interaction interface diagram and Figure 10The human-computer interface display communication architecture diagram, the embodiment details the display communication architecture of the AI algorithm controller (151) and the human-computer interaction interface (152) in the core control unit (150), and the hardware configuration, partition display logic of the human-computer interaction interface (152) in the "close interaction function (only one-way display)" state, so as to ensure that the technical personnel in the technical field can completely reproduce the technical scheme of "multi-source data acquisition -> AI processing -> interface display".

[0257] I. Human-computer interaction display communication architecture (signal flow logic)

[0258] As shown in Figure 10 , the multi-source external device, the AI algorithm controller (151), and the human-computer interaction interface (152) realize data interaction through the link of "signal input -> AI processing -> display output", and the communication mode and function of each link are as follows:

[0259] Signal input of multi-source external device: various external devices transmit signals to the AI algorithm controller (151) through the RS485 bus (partially adopting the Modbus-RTU protocol):

[0260] Concentration detection: the PID concentration instrument (122) transmits the waste gas concentration analog signal (4-20mA) every 2 seconds, which is converted into a digital quantity by the built-in A / D conversion module of the AI algorithm controller (151);

[0261] Energy consumption metering: the intelligent gas meter (F1010, not all of the present application) and the intelligent electric energy meter (F1020, not all of the present application) respectively transmit digital signals (Modbus-RTU protocol) of gas and electric energy consumption;

[0262] Temperature acquisition: the PT100 temperature sensor (114-1, built-in in the condenser) transmits the condenser temperature analog signal (4-20mA, through signal conditioning circuit + A / D conversion), and the DS18B20 temperature sensor (1020) outside the device transmits the environmental temperature digital signal every 30 seconds;

[0263] Device state and historical data: the device power supply circuit (1010) transmits the on-off state digital signal (dry contact level, 0 = off / 1 = on), and the zeolite runner PLC controller (F140, not all of the present application) transmits the total boot time and runner speed digital signal (Modbus-RTU protocol).

[0264] Processing and transmission of AI algorithm controller: after receiving multi-source signals, the AI algorithm controller (151) performs "preprocessing -> autonomous calculation -> standardized format conversion":

[0265] Preprocessing: A / D conversion is performed on analog signals, and Modbus-RTU protocol analysis is performed on digital signals;

[0266] Autonomous computing: based on the pre-processed data, calculate the heating energy saving ratio, electricity energy saving ratio, total energy saving ratio, total cost saving, etc. (such as total energy saving ratio = (heating energy saving ratio + electricity energy saving ratio) / 2) ;

[0267] Format conversion: convert the calculation results and the original signals that need to be directly displayed (such as real-time concentration, temperature) into standardized display data format according to the display rules preset by the human-computer interaction interface (such as numerical accuracy, unit identification) ;

[0268] Transmission: through RS485 bus (Modbus-RTU protocol, baud rate 9600bps), standardized data is transmitted to the display driving module (152-1) of the human-computer interaction interface (152).

[0269] The display driving of the human-computer interaction interface is realized through "module cooperation" :

[0270] Display driving module (152-1) : receive standardized data transmitted by AI algorithm controller, analyze data type (such as "total energy saving ratio" "real-time concentration"), and drive the corresponding function area of the screen to display;

[0271] Touch screen RTC module (152-2) : generate real-time time signal (format: YYYY-MM-DDHH:mm:ss) every 1 second, and transmit it to the touch screen display module (152-3) through the internal bus;

[0272] Touch screen display module (152-3) : integrate the parsed data of the display driving module and the time signal of the RTC module, and complete the integrated display of the whole interface information.

[0273] II. Partition display logic of human-computer interaction interface (general + consistency of closed interaction)

[0274] The human-computer interaction interface (152) adopts a 7-inch capacitive touch screen, which is divided into time display area, cumulative statistics area, energy consumption comparison area, real-time parameter area, cost statistics area and equipment running state display area according to functions. This part explains the general display logic of "opening interaction function" and the consistency of "display content update logic" in the special state of "closing interaction function (only one-way display)" (the disablement of interaction function is described in section III).

[0275] (I) General hardware and communication basis

[0276] The touch screen has the hardware capabilities of touch operation, parameter input and interface switching, and communicates with the AI algorithm controller (151) through RS485 interface with "baud rate 9600bps, Modbus-RTU protocol".

[0277] (ii) Area display logic (data source, calculation and update rule)

[0278] Whether the interaction function is turned on or not, the data source, calculation logic and update period of each area remain the same, as follows:

[0279] The time display area (910) displays the format: "YYYY year MM month DD day, HH:mm:ss, X day" (example: "2025 year 7 month 25 day, 10:30:33, Friday"). The data source: the touch screen RTC module (152-2) generates a real-time time signal every 1 second, which is transmitted to the touch screen display module (152-3) through the internal bus to drive the time display area (910) to automatically refresh.

[0280] Cumulative statistics area (920, total boot time, total energy saving ratio, total energy saving time)

[0281] Total boot time (921): The total running time of the zeolite rotary system accumulated by the zeolite rotary PLC controller (F140), which is converted to "XX hours XX minutes XX seconds" format by the AI algorithm controller (151), and then transmitted to the display driving module (152-1) through RS485 (Modbus-RTU protocol). After the display driving module (152-1) analyzes the signal, it drives the total boot time (921) area of the touch screen display module (152-3) to complete the display, which is automatically updated every 10 minutes.

[0282] Total energy saving ratio (922): The AI algorithm controller (151) calculates the arithmetic mean of "heating energy saving ratio (933)" and "electric energy saving ratio (936)" (formula: total energy saving ratio = (heating energy saving ratio + electric energy saving ratio) / 2), which is converted to "XX.X%" format and then transmitted to the display driving module (152-1) through RS485. After the display driving module (152-1) analyzes the signal, it drives the total energy saving ratio (922) area of the touch screen display module (152-3) to complete the display, which is updated every 10 minutes.

[0283] Total energy saving time (923): The AI algorithm controller (151) accumulates the duration of the energy saving adjustment instruction sent by the device to the zeolite rotary system (F100), which is converted to "XX hours XX minutes XX seconds" format and then transmitted to the display driving module (152-1) through RS485. After the display driving module (152-1) analyzes the signal, it drives the total energy saving time (923) area of the touch screen display module (152-3) to complete the display, which is automatically updated every 10 minutes.

[0284] The energy consumption comparison area (930, updated at every hour) includes six sub-areas: "unit time energy-saving furnace heating consumption (931), unit time non-energy-saving furnace heating consumption (932), heating energy-saving ratio (933), unit time energy-saving device power consumption (934), unit time non-energy-saving device power consumption (935), and electric energy-saving ratio (936)".

[0285] Unit time energy-saving furnace heating consumption (931): The intelligent gas meter (F1010) collects the gas consumption data in the "energy-saving state (the device is turned on for energy-saving control)", and transmits it to the AI algorithm controller (151) through RS485. The controller converts the data into "XX.Xm 3 / h" format, and then transmits it to the display driving module (152-1) through RS485 (Modbus-RTU protocol). After the display driving module (152-1) analyzes the signal, it drives the unit time energy-saving furnace heating consumption (931) area of the touch screen display module (152-3) to complete the display.

[0286] Unit time non-energy-saving furnace heating consumption (932): The intelligent gas meter (F1010) collects the gas consumption data in the "non-energy-saving state (the device is turned off for energy-saving control)", and transmits it to the AI algorithm controller (151) through RS485 (Modbus-RTU protocol). The controller converts the data into "XX.Xm 3 / h" format, and then transmits it to the display driving module (152-1) through RS485. After the display driving module (152-1) analyzes the signal, it drives the unit time non-energy-saving furnace heating consumption (932) area of the touch screen display module (152-3) to complete the display.

[0287] Heating energy-saving ratio (933): The AI algorithm controller (151) calculates it by the formula:

[0288] Heating energy-saving ratio = (unit time non-energy-saving furnace heating consumption - unit time energy-saving furnace heating consumption) / unit time non-energy-saving furnace heating consumption x 100%

[0289] After the calculation result is converted into "XX.X%" format, it is transmitted to the display driving module (152-1). After the display driving module (152-1) analyzes the signal, it drives the heating energy-saving ratio (933) area of the touch screen display module (152-3) to complete the display.

[0290] Unit time energy saving equipment power consumption (934): the intelligent power meter (F1020) collects the power consumption data of the "energy saving state", and transmits it to the AI algorithm controller (151) through RS485; the controller converts the data into "XX.XkW·h" format, and then transmits it to the display driving module (152-1) through RS485 (Modbus-RTU protocol); after the display driving module (152-1) analyzes the signal, the unit time energy saving equipment power consumption (934) area of the touch screen display module (152-3) is driven to complete the display.

[0291] Unit time non-energy saving equipment power consumption (935): the intelligent power meter collects the power consumption data of the "non-energy saving state", and transmits it to the AI algorithm controller (151) through RS485 (Modbus-RTU protocol); the controller converts the data into "XX.XkW·h" format, and then transmits it to the display driving module (152-1) through RS485; after the display driving module (152-1) analyzes the signal, the unit time non-energy saving equipment power consumption (935) area of the touch screen display module (152-3) is driven to complete the display.

[0292] Electric energy saving ratio (936): the AI algorithm controller (151) calculates by formula:

[0293] Electric energy saving ratio = (unit time non-energy saving equipment power consumption-unit time energy saving equipment power consumption) / unit time non-energy saving equipment power consumption × 100%

[0294] After the calculation result is converted into "XX.X%" format, it is transmitted to the display driving module (152-1); after the display driving module (152-1) analyzes the signal, the electric energy saving ratio (936) area of the touch screen display module (152-3) is driven to complete the display.

[0295] Real-time parameter area (940)

[0296] Real-time concentration value (941): the concentration detection unit (120) PID concentration instrument (122) transmits the waste gas concentration analog signal (4-20mA) once every 2 seconds, which is converted into digital quantity by the built-in A / D conversion module of the AI algorithm controller (151); after the AI algorithm controller (151) standardizes the signal into "XX.Xppm" format (display precision: 1 bit after decimal point), it is transmitted to the display driving module (152-1) through RS485; after the display driving module (152-1) analyzes the signal, the real-time concentration value (941) area of the touch screen display module (152-3) is driven to complete the display.

[0297] Condenser temperature (942): The PT100 temperature sensor (114-1, range: 0-100℃, accuracy ±0.5℃) built-in the condenser first collects the real-time temperature signal of the condenser; the original temperature signal is transmitted to the signal conditioning circuit of the pre-filtering unit (110) and converted into a 4-20mA standard analog signal after "amplification and filtering" processing; then, the 4-20mA analog signal is transmitted to the AI algorithm controller (151), which is converted into a digital signal by the A / D conversion module built-in the AI algorithm controller (151); after the AI algorithm controller (151) formats the digital signal, it is forwarded to the display driver module (152-1) of the human-computer interaction interface (152); after the display driver module (152-1) analyzes the signal, it drives the condenser temperature (942) area of the touch screen display module (152-3) to complete the display, with an update cycle of every 10 seconds and a display format of "XX℃" (example: "9℃").

[0298] Ambient temperature (943): The DS18B20 temperature sensor (1020, an extension component of the device, range: -20-60℃, accuracy ±0.2℃) outside the device is directly transmitted to the display driver module (152-1) every 30 seconds; after the display driver module (152-1) analyzes the signal, it drives the ambient temperature (943) area of the touch screen display module (152-3) to complete the display, with a format of "XX.X℃".

[0299] Cost statistics area (950)

[0300] Total savings (951): The AI algorithm controller (151) calculates the total daily energy-saving income (based on the local real-time electricity and gas prices, combined with the "hourly energy consumption difference" to calculate the daily income) from Monday to Sunday in the "this week's energy-saving cost statistics bar chart (952)", converts it into "XXXXX.XX yuan" format (accurate to two decimal places), and then transmits it to the display driver module (152-1) through RS485; after the display driver module (152-1) analyzes the signal, it drives the total savings (951) area of the cost statistics area (950) of the touch screen display module (152-3) to complete the display.

[0301] This week's energy-saving cost statistics bar chart (952): The AI algorithm controller (151) automatically calculates the daily energy-saving income every day at 0 o'clock and pushes the data to the display driver module (152-1); the display driver module (152-1) generates a bar chart based on the data with the horizontal axis as Monday to Sunday (date) and the vertical axis as cost (unit: yuan), and after the display driver module (152-1) analyzes the signal, it drives the this week's energy-saving cost statistics bar chart (952) area of the cost statistics area (950) of the touch screen display module (152-3) to complete the display.

[0302] Device running state display area (960, ≤1 second delay real-time display) display logic: indicator light is "round state light", green represents "running", red represents "off". When the display driving module (152-1) receives the "green running" signal sent by the AI algorithm controller (151), the indicator light displays green; when the "red off" signal is received, the indicator light displays red. State determination and data source:

[0303] Energy-saving start state indicator light (961): The AI algorithm controller (151) determines according to "whether to send energy-saving adjustment instructions to the zeolite rotary system" - if in the instruction sending state (energy-saving control is turned on), send a "green running" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the energy-saving khaki state indicator light (961) of the device running state display area (960) to display green; if no instruction is sent (energy-saving control is turned off), send a "red off" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the energy-saving khaki state indicator light (961) of the device running state display area (960) to display red.

[0304] Device start state indicator light (962): The AI algorithm controller (151) determines according to "the on-off state of the power supply circuit of the energy-saving control device" - if the power supply circuit is on (the device is powered on and running), send a "green running" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the device start state indicator light (962) of the device running state display area (960) to display green; if the power supply circuit is off (the device is powered off), send a "red off" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the device start state indicator light (962) of the device running state display area (960) to display red.

[0305] Peristaltic pump start state indicator light (963): The AI algorithm controller (151) determines according to "the on-off state of the power supply circuit of the peristaltic pump (115)" - if the power supply circuit is on (the peristaltic pump is running), send a "green running" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the peristaltic pump start state indicator light (963) of the device running state display area (960) to display green; if the power supply circuit is off (the peristaltic pump is stopped), send a "red off" signal to the display driving module (152-1), and after the display driving module (152-1) analyzes the signal, drive the peristaltic pump start state indicator light (963) of the device running state display area (960) to display red.

[0306] Condenser open state indicator light (964): AI algorithm controller (151) determines according to the "condenser (114) power supply circuit on-off state" judgment - if the power supply circuit is on (condenser operation), send "green operation" signal to display drive module (152-1), display drive module (152-1) after analyzing the signal, drive the condenser open state indicator light (964) of the equipment running state display area (960) to display green; if the power supply circuit is disconnected (condenser stop), send "red off" signal to display drive module (152-1), display drive module (152-1) after analyzing the signal, drive the condenser open state indicator light (964) of the equipment running state display area (960) to display red.

[0307] Three, "close the interactive function" of the special state

[0308] In this state, the touch response function of the touch screen is disabled by software (specifically, the touch input response is closed by the QTouchDevice: setEnabled (false) interface of the Qt framework in the embedded system), all physical touch operations (such as clicking, long pressing, etc.) do not trigger parameter setting, mode switching and other functions.

[0309] The interface only serves as a data display terminal, and the update logic of the display content in each area is completely consistent with that in the "open interactive function" state (such as the data source, calculation rule, and update period remain unchanged), and only the user operation response link is removed to prevent misoperation of on-site personnel from causing abnormal system parameters and ensure the stability of equipment operation.

[0310] It should be noted that the above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as limiting the scope of the present patent. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the protection scope of the present patent should be subject to the appended claims.

Claims

1. An energy-saving control device applied to a zeolite rotor system, characterized in that, include: The gas pretreatment unit includes a filter 1, an intake pump, a filter 2, a condenser, and a peristaltic pump connected in series along the waste gas treatment flow direction. The filter 1 is used to initially intercept particulate impurities in the waste gas. The intake pump is used to extract the industrial waste gas filtered by the filter 1 and drive the system airflow circulation. The filter 2 is used to secondarily intercept residual particulate impurities in the waste gas. The condenser is used to reduce the humidity of the waste gas. The peristaltic pump is used to pump out the wastewater generated during the waste gas pretreatment process. The concentration detection unit includes a PID concentration meter and a flow meter; the flow meter is used to control the flow rate of the exhaust gas sample entering the PID concentration meter, the PID concentration meter is used to detect the concentration value of the target pollutant in the exhaust gas sample in real time and output a concentration signal, and the flow meter synchronously outputs a real-time flow signal. A status feedback unit is connected to the PLC controller built into the zeolite rotor system and is used to receive the actual rotor speed signal and actual desorption temperature signal of the zeolite rotor system transmitted in real time by the PLC controller. Control unit, the control unit including an AI algorithm controller; The AI ​​algorithm controller is connected to the concentration detection unit and the status feedback unit respectively, and is used to receive the concentration signal, flow signal, actual rotor speed signal and actual desorption temperature signal. The AI ​​algorithm controller has a built-in preset machine learning algorithm. The algorithm uses historical exhaust gas concentration, flow rate, rotor speed, desorption temperature and corresponding energy consumption data as training samples, and takes "minimizing equipment energy consumption when exhaust gas treatment efficiency is ≥90%" as the optimization goal. It generates rotor speed adjustment command and desorption temperature adjustment command adapted to the current exhaust gas conditions through real-time calculation, and transmits the adjustment command to the PLC controller built into the zeolite rotor system. The PLC controller built into the zeolite rotor system is connected to the AI ​​algorithm controller and the actuators (rotor drive motor and desorption heating device) of the zeolite rotor system. It is used to receive the speed adjustment command and the desorption temperature adjustment command, and convert the command into an execution signal to control the actuator to adjust the rotor speed and desorption temperature. The status feedback unit transmits the received actual rotor speed signal and actual desorption temperature signal back to the AI ​​algorithm controller in real time, forming an energy-saving control closed loop of "device detection and calculation - zeolite rotor system built-in PLC execution - device feedback optimization".

2. The device as described in claim 1, characterized in that, Both filter 1 and filter 2 are polymer filters made of PE material.

3. The device as described in claim 2, characterized in that, The PE polymer filter has a filtration accuracy of 1-10μm.

4. The device as described in claim 1, characterized in that, The peristaltic pump has a pumping flow rate range of 0.1-2L / h and is equipped with anti-clogging and shutdown protection functions. It is used to stably pump out sewage and liquid impurities condensed in the exhaust gas generated by the condenser.

5. The device as described in claim 1, characterized in that, The condenser uses closed-loop temperature control to set the condensation temperature to 5-10℃ and stably reduce the relative humidity of the exhaust gas to ≤30%.

6. The device as described in claim 1, characterized in that, The PID concentration meter has a detection accuracy of ±5%FS (full scale), a response time of ≤10s to the target pollutant, and an automatic zero-point calibration function.

7. The device as described in claim 1, characterized in that, The flow meter is a rotor flow meter, and its flow control range is 0.5-5m³. 3 / h, with a flow control accuracy of ±2.5%FS.

8. The device as described in claim 1, characterized in that, The filter 1, suction pump, filter 2, condenser, peristaltic pump and flow meter are all connected by a VOCs-resistant hose with an inner diameter of 10-15 mm.

9. The device as described in claim 8, characterized in that, The VOCs-resistant hoses are connected by tees, bends, and flow-limiting valves.

10. The device as claimed in claim 1, characterized in that, The machine learning algorithm is the BP neural network algorithm.

11. The device as claimed in claim 10, characterized in that, The network structure of the BP neural network algorithm is as follows: the input layer contains 4 neurons, corresponding to the concentration value, flow rate value, actual rotor speed, and actual desorption temperature, respectively; the output layer contains 2 neurons, corresponding to the rotor speed adjustment value and the desorption temperature adjustment value, respectively; the hidden layer is set to 2 layers, each containing 12 neurons, and the hidden layer uses the Sigmoid activation function, while the output layer uses the linear activation function.

12. The device as claimed in claim 1, characterized in that, The control unit also includes a human-machine interface, which is a touch screen; the touch screen is used to display the following information: total uptime of the zeolite rotor equipment, total energy saving ratio of the system, and total energy saving time. Real-time concentration values, condenser operating temperature values, and ambient temperature values; power consumption and gas consumption of the zeolite rotor equipment per unit time in energy-saving mode (equipment on) and non-energy-saving mode (equipment off); operating / shutdown status of the zeolite rotor equipment, this energy-saving control equipment, peristaltic pump, and condenser; bar chart of daily cost savings for this week and status bar of total cost savings for this week.

13. The device as claimed in claim 1, characterized in that, The AI ​​algorithm controller achieves integrated signal connection with the PLC controller built into the zeolite rotor system through an RS485 communication interface, or achieves remote signal connection with the PLC controller through an industrial Ethernet (Profinet protocol), with a communication rate ≥9600bps.

14. The device as claimed in claim 1, characterized in that, It also includes a furnace temperature monitoring unit; The furnace temperature monitoring unit is connected to the PLC controller built into the zeolite rotor system to receive the real-time temperature signal transmitted by the PLC controller (the temperature signal is collected by the PT100 platinum resistance temperature sensor of the furnace matched with the zeolite rotor, and the furnace includes a CO catalytic furnace or an RTO combustion furnace), and transmits the temperature signal to the AI ​​algorithm controller.

15. The device as claimed in claim 14, characterized in that, The AI ​​algorithm controller is configured as follows: The temperature signal is subjected to multi-dimensional feature extraction, and the extracted content includes the furnace type identifier and real-time temperature value; Based on the preset temperature threshold matching rules for different types, it is determined whether the temperature parameters of the currently monitored furnace meet the corresponding safe operating range: if it is a CO catalytic furnace, it is determined whether its temperature is in the range of 280℃≤temperature≤580℃; if it is an RTO combustion furnace, it is determined whether its temperature is in the range of 760℃≤temperature≤900℃.

16. The device as claimed in claim 15, characterized in that, The AI ​​algorithm controller is also configured to: Based on the temperature determination result, a normal operation signal or an abnormal operation alarm signal is generated, and the alarm signal is fed back to the PLC controller built into the zeolite rotor system; The normal operation signal includes a real-time temperature value, and the abnormal operation alarm signal includes the temperature anomaly type (exceeding the upper or lower limit) and the real-time temperature value. The PLC controller integrated with the zeolite rotor system is configured to perform safety protection actions according to the preset safety protection rules of the zeolite rotor system based on the received alarm signal.

17. The device as claimed in claim 14, characterized in that, The AI ​​algorithm controller is also configured to perform furnace temperature control functions: The AI ​​algorithm controller is connected to the PLC controller built into the zeolite rotor system. It optimizes the furnace temperature through multiple algorithms, generates a furnace temperature adjustment signal, and transmits it to the PLC controller. The PLC controller then controls the furnace heating device (such as a gas proportional valve) to achieve dynamic adjustment of the furnace temperature.

18. The device as claimed in claim 17, characterized in that, When performing furnace temperature control, the AI ​​algorithm controller predicts the basic furnace temperature using a BP neural network algorithm. Using historical desorption temperatures, corresponding furnace temperatures, and associated energy consumption data (such as electricity and gas) and waste gas treatment efficiency data as training samples, a mapping model between desorption temperature and furnace temperature is established, and basic furnace temperature parameters adapted to current desorption requirements are output.

19. The device as claimed in claim 18, characterized in that, When executing the furnace temperature control function, the AI ​​algorithm controller dynamically optimizes the furnace temperature correction range through reinforcement learning algorithm: Using the desorption ratio (the ratio of desorption air volume to treatment air volume) as the decision variable, and the "energy consumption reduction rate" and "treatment efficiency compliance rate" as the reward function, the optimal desorption ratio is dynamically output through interactive learning with real-time operating conditions (fluctuations in exhaust gas concentration and changes in flow rate), and the correction range for the furnace temperature is calculated based on the desorption ratio.

20. The device as claimed in claim 19, characterized in that, When executing the furnace temperature control function, the AI ​​algorithm controller introduces an MPC (Model Predictive Control) predictive model to achieve collaborative optimization. By integrating the basic furnace temperature parameters output by the BP neural network and the temperature correction amplitude output by the reinforcement learning algorithm, and combining the exhaust gas conditions (concentration and flow trend) predicted by the time-series prediction model for the next 5-10 minutes, the most suitable furnace temperature adjustment signal adapted to the real-time operating conditions is finally generated.

21. The device as claimed in claim 20, characterized in that, The input source signals of the time-series prediction model include: the real-time concentration signal output by the concentration detection unit, the real-time flow signal output by the flow meter, the future production schedule provided by the production planning system (derived from the target customer's own production plan), and the past operating conditions stored in the historical database. The AI ​​algorithm controller extracts time-series features, fits trends, and matches patterns from the aforementioned multi-source data using the time-series prediction model, predicts the exhaust gas conditions (concentration and flow trends) for the next 5-10 minutes, and uses the prediction results for the collaborative optimization of the MPC prediction model.

Images