Anti-condensation temperature and humidity regulation and control system and method for remote deep learning prediction aided decision-making

Abstract

The invention provides an anti-condensation temperature and humidity regulation and control system and method for remote deep learning prediction aided decision making, and the method is characterized in that a field controller is connected with a temperature and humidity collection device, obtains the environment temperature and humidity, calculates the dew point temperature, obtains the surface temperature of a protected part, and starts and stops a heating and / or dehumidification device by comparing the difference between the two values with a preset threshold value; the temperature and humidity acquisition device acquires field environment temperature and humidity data for the field controller to calculate the dew point temperature, and the temperature and humidity acquisition device is also used for acquiring the surface temperature of the protected part; the heating device is used for raising environment temperature; the dehumidification device is used for reducing the environment humidity; the communication module is used for data transmission between the field controller and the remote prediction terminal, sending field data and receiving prediction information; the remote prediction terminal obtains field data through the communication module; an anti-condensation system structure combining local autonomous control and remote prediction assistance is established, and redundancy fault-tolerant capability is achieved; even if the remote prediction function fails, the field control can still operate independently and reliably.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of anti-condensation temperature and humidity control, in particular to an anti-condensation temperature and humidity control system and method assisted by remote deep learning prediction decision-making. BACKGROUND In many industrial fields and places with strict environmental conditions, such as substations, communication base stations, industrial control cabinets, smart grid equipment, etc., maintaining a suitable temperature and humidity environment is a key factor to ensure the stable operation of equipment, prolong the service life of equipment, and ensure the safety of the system. Once the environmental temperature and humidity is out of control, condensation phenomenon is prone to occur, which may cause a series of serious problems. Taking substations as an example, a large number of electrical equipment are arranged in a centralized manner, and the internal space is relatively closed. Due to insufficient sealing and underground moisture return, moisture easily enters, especially at the cable trench, the relative humidity is often in a high interval of 70%-95%, and these moist air can enter the switch cabinet through the gap at the bottom of the switch cabinet. At the same time, the bus duct in the substation is connected with the outdoor, which is easy to form a chimney effect, causing the warm and humid air to gather in the bus duct, and the humidity is usually 50%-70%, and the humidity will be higher in humid weather. When the environmental humidity is high and the temperature difference is large, the moisture in the air is prone to condense into dew in the cabinet, making the condensation phenomenon at the top of the bus duct more prominent. Long-term exposure to such condensation environment will reduce the insulation strength of the live parts, and may cause partial discharge, which brings great safety hazards to the operation of electrical equipment. In communication base stations, various communication equipment is extremely sensitive to temperature and humidity. Once condensation occurs, electronic components may be damaged due to short circuit, resulting in communication interruption, affecting signal transmission quality and stability, and seriously affecting communication services. In industrial control cabinets, there are many internal components and the space is compact. When the environmental temperature is high and the temperature difference is large, condensation is also prone to occur in the cabinet, which may cause equipment aging, insulation strength reduction, secondary terminal breakdown, material mildew, steel structure corrosion, and other problems, eventually leading to safety accidents. At present, the traditional methods used to solve the condensation problem have many drawbacks. The common heating type dehumidifier reduces the relative humidity by increasing the temperature in the cabinet, but it does not reduce the water content in the air in the cabinet. Once the environmental temperature changes, the humid air will quickly condense, which cannot fundamentally solve the problem. The fan type dehumidifier relies on air convection to exchange with dry air outside to reduce indoor humidity. However, when the environmental humidity is high, it cannot effectively solve the condensation problem, and it is also easy to make dust and dirt enter the cabinet, causing additional damage to the equipment. In addition, some existing anti-condensation solutions often have single logic, cannot realize intelligent control, have high energy consumption and are not environmentally friendly, and are difficult to meet the needs of complex and variable actual application scenarios. There are also some solutions that only control the temperature and humidity parameters independently for a single device, such as installing a ventilation system in the cable semi-layer, installing an industrial dehumidifier indoors, and installing a heater in the switch cabinet. However, due to the coupling and mutual influence of temperature and humidity, especially the real-time change of humidity with temperature, this single control method has little effect on the improvement of the overall electrical equipment operating environment, and cannot comprehensively and effectively solve the condensation problem.

[0002] Therefore, it is urgent to develop a remote deep learning prediction assisted decision-making anti-condensation temperature and humidity regulation system and method that can comprehensively consider temperature and humidity changes, achieve intelligent and accurate control, have high energy efficiency, and can still operate stably when remote assistance fails, to meet the strict requirements of various industries for the reliability and stability of equipment operating environment. SUMMARY

[0003] The main purpose of the present application is to provide a remote deep learning prediction assisted decision-making anti-condensation temperature and humidity regulation system and method, which solves the problem that the prior art lacks an anti-condensation system that can both operate stably locally and utilize remote prediction information for optimized control.

[0004] To solve the above technical problems, the technical solution adopted by the present application is: a remote deep learning prediction assisted decision-making anti-condensation temperature and humidity regulation system, comprising a field controller, a temperature and humidity acquisition device, a heating device, a dehumidifying device, a communication module, and a remote prediction terminal; The field controller connects the temperature and humidity acquisition device to obtain the environmental temperature and humidity and calculate the dew point temperature, and simultaneously obtains the surface temperature of the protected part, and starts or stops the heating and / or dehumidifying device by comparing the difference between the two with a preset threshold value; The temperature and humidity acquisition device acquires field environmental temperature and humidity data for the field controller to calculate the dew point temperature, wherein the temperature and humidity acquisition device is also used to acquire the surface temperature of the protected part; The heating device is used to raise the environmental temperature; The dehumidifying device is used to reduce the environmental humidity; The communication module is used for data transmission between the field controller and the remote prediction terminal, to send field data and receive prediction information; The remote prediction terminal obtains field data through the communication module; In a preferred solution, the method comprises: S1, acquiring field environmental temperature and humidity data, calculating the dew point temperature through the field controller, and simultaneously acquiring the surface temperature of the protected part; S2, judging the condensation risk according to the difference between the surface temperature and the dew point temperature, and if the difference is lower than a preset safety threshold value, starting the heating device or the dehumidifying device through the field controller to adjust the environmental conditions; S3, transmitting the field data to the remote prediction terminal through the communication module, and receiving the environmental change prediction information of the remote prediction terminal; S4, predicting the change trend of the environmental temperature and humidity in a future predetermined period of time based on a deep learning model through the remote prediction terminal, and sending the prediction information to the field controller; S5, executing the regulation and control decision through the field controller by synthesizing the prediction information and real-time data, wherein the local real-time control is preferentially executed, and the prediction information is used as an auxiliary optimization decision.

[0005] In the preferred embodiment, the condensation risk is determined according to the difference between the surface temperature and the dew point temperature in step S2, comprising: Periodically acquiring the environmental temperature and humidity data through the field controller; Calculating the dew point temperature of the current environment according to the environmental temperature and humidity data, wherein the dew point temperature calculation uses the formula , wherein f is the dew point temperature based on the environmental temperature and the relative humidity RH; Acquiring the surface temperature of the protected part, and calculating the difference between the surface temperature and the dew point temperature, wherein the difference formula is: , wherein is the surface temperature, is the dew point temperature; If the difference is lower than a preset safety threshold, it is determined that there is a condensation risk, and the heating device or the dehumidification device is triggered to start; During the operation of the heating device or the dehumidification device, the difference is continuously monitored until the difference is higher than another preset threshold, and the corresponding device is stopped.

[0006] In the preferred embodiment, the heating device or the dehumidification device is started to adjust the environmental conditions through the field controller in step S2, comprising: Determining the dominant factor of the environmental temperature or humidity according to the source of the condensation risk; If the condensation risk is mainly caused by the low environmental temperature, the heating device is preferentially started to increase the temperature through the field controller; If the condensation risk is mainly caused by the high environmental humidity, the dehumidification device is preferentially started to reduce the humidity through the field controller; If the condensation risk is affected by both temperature and humidity, the heating device and the dehumidification device are simultaneously started to quickly adjust the environmental conditions.

[0007] In the preferred embodiment, the field data is transmitted to the remote prediction terminal through the communication module in step S3, comprising: Collecting field environment temperature, humidity, dew point temperature, surface temperature and equipment running state data periodically through the communication module; Transmitting field data to remote prediction terminal for storage and analysis; Receiving environmental change prediction information generated by the remote prediction terminal based on field data through the communication module; Feeding back the prediction information to the field controller for subsequent decision assistance; Among them, the communication module adopts a 4G cellular network module, and the data transmission period is every 1-2 minutes.

[0008] In the preferred scheme, in step S4, the remote prediction terminal predicts the change trend of the environment temperature and humidity in the future predetermined period of time based on the deep learning model, including: Obtaining field data as input through the remote prediction terminal; Inputting the field data into the pre-trained deep learning model for time series analysis; Outputting the environmental temperature and humidity change trend data in the future predetermined period of time through the deep learning model, and the deep learning model can adopt the algorithm structure suitable for time series prediction of long short-term memory network LSTM and time series convolution network TCN; Outputting the environmental temperature and humidity change trend data in the future predetermined period of time through the deep learning model, for example, predicting the expected temperature and humidity every 5 minutes in the next 1 hour ; ; Generating prediction information according to the change trend data and sending it to the field controller through the communication module; In the preferred scheme, in step S5, the field controller executes control decisions by comprehensively analyzing the prediction information and real-time data, including: Receiving prediction information through the field controller; Comprehensively analyzing the prediction information and the current real-time data; If the prediction information indicates that the future condensation risk increases, the heating device or dehumidification device is started in advance for preventive adjustment; If the prediction information indicates that the environmental conditions will improve soon, the heating device or dehumidification device is started later to reduce unnecessary operation; In any case, the safety control logic is preferentially executed according to the difference of real-time data; In the preferred scheme, the specific steps of steps S4-S5 are: The field data includes time series form of environmental temperature , relative humidity , dew point temperature , surface temperature of key parts and running state mark of heating / dehumidification device After data cleaning, the missing values are filled by cubic spline interpolation, and then the sliding window statistics, temperature and humidity change rates, and the cumulative amount of dew point and surface temperature difference are extracted to splice into high-dimensional input vectors ; The input vectors are input into the pre-trained improved LSTM model (with attention mechanism) for time series analysis. The model input is a historical data window containing time steps. The attention layer is used to calculate the weight of each historical time step. The root mean square error is used as the loss function for training. After multi-scale residual correction, the predicted values of ambient temperature and humidity every 5 minutes in the next hour are output and respectively. The 95% confidence interval is estimated based on the Bootstrap method . The future dew point temperature is calculated using the improved Magnus-Tetens formula based on the predicted values, and the condensation risk index is defined as , where is the safety threshold, is the danger threshold. The prediction information containing predicted values, risk index, and confidence interval is sent to the field controller through the communication module , where is the dew point temperature at each future time step, is the condensation risk index . After receiving the prediction information, the field controller first parses the predicted values of ambient temperature , humidity , dew point temperature at each future time step, condensation risk index and confidence interval, and calculates the credibility score C of the predicted data: The credibility score C is calculated as , where is the variance of temperature prediction value, is the reference variance . A multi-objective optimization model is constructed with the objectives of "minimizing condensation risk" and "minimizing device operating energy consumption". The decision variable is defined, and the objective function is , where is the weight, is the energy consumption function . The optimal strategy is solved based on dynamic programming , combined with real-time temperature difference for fusion correction. If If not, the device is forced to start, otherwise, the start time is determined according to the optimal action and the prediction confidence, and the prediction model is corrected online based on real-time data, and finally the device control signal is output.

[0009] In the preferred solution, in step S5, the local real-time control is preferentially executed, including: When the remote prediction terminal or the communication module fails, the anti-condensation control is independently executed by the field controller based on real-time data; The field controller continuously monitors the difference between the surface temperature and the dew point temperature; If the difference is below the preset safety threshold, the heating device or the dehumidification device is immediately started; The field controller maintains the local control logic to ensure that the system still operates stably when the remote assistance fails.

[0010] In the preferred solution, when the remote prediction terminal or the communication module fails, the anti-condensation control is independently executed by the field controller based on real-time data, the field controller continuously obtains the environmental temperature and humidity data through the temperature and humidity collection device, calculates the dew point temperature, and continuously monitors the difference between the surface temperature of the protected part obtained through the surface temperature sensor and the dew point temperature; The field controller has preset start and stop thresholds, when the difference is below the start threshold, the heating device or the dehumidification device is immediately started, wherein the dominant factor of condensation risk is determined by analyzing the deviation of the environmental temperature from the normal working temperature range of the device and the deviation of the environmental humidity from the safe humidity range, the corresponding device is preferentially started, and if both the temperature and the humidity are out of range, both devices are started; After the heating device and the dehumidification device are started, the field controller executes the device protection logic and sets the minimum continuous running time, even if the difference rises above the start threshold during the running, the device needs to run to the end of the minimum continuous running time, when the running time exceeds the minimum continuous running time, the field controller compares the difference with the stop threshold, if the difference continuously exceeds the stop threshold and remains for a preset time, a stop instruction is issued, otherwise the running continues; The field controller records the time, running time, and difference change curve data of each start and stop and stores them in the local cache, and monitors the running state of the heating device and the dehumidification device, if a device continuously runs for more than a preset maximum single running time limit or starts and stops more than a preset number of times within a preset time, a warning is issued through the local indicator light; After the remote prediction terminal or the communication module recovers from the failure, the field controller uploads the local cache data to the remote prediction terminal through the communication module, and in the whole process, the field controller maintains the above-mentioned local control logic to ensure that the system still operates stably when the remote assistance fails.

[0011] The application provides a remote deep learning prediction assisted decision-making anti-condensation temperature and humidity regulation system and method, establishes an anti-condensation system structure combining local autonomous control and remote prediction assistance, and has redundancy fault tolerance capability; even if the remote prediction function fails, the on-site control can still operate independently and reliably, ensuring the safety and stability of the system.

[0012] The environment trend is predicted by using deep learning, the control decision has foresight, and the condensation can be prevented in advance or unnecessary operation can be reduced. Compared with traditional pure on-site control, the frequent start and stop of the heater and dehumidifier are significantly reduced, and the energy consumption and equipment wear are reduced.

[0013] The reliability of local control and the optimization capability of remote intelligence are combined, the traditional anti-condensation device is upgraded to an intelligent anti-condensation system, and significant improvement is achieved in reliability, energy efficiency and intelligent level. BRIEF DESCRIPTION OF DRAWINGS

[0014] The application will be further described below in combination with the drawings and examples: Fig. 1 is a structural block diagram of the anti-condensation temperature and humidity regulation system of the application; Fig. 2 is a local control flowchart of the application; Fig. 3 is a remote prediction assisted decision-making flowchart of the application. DETAILED DESCRIPTION

[0015] Example 1 As shown in Figs. 1-3 , an anti-condensation temperature and humidity regulation system with remote deep learning prediction assisted decision-making includes a local controller, a temperature and humidity collection device, a heating device, a dehumidification device, a communication module and a remote prediction terminal; The local controller connects the temperature and humidity collection device to obtain the environmental temperature and humidity and calculate the dew point temperature, and at the same time obtains the surface temperature of the protected part, and starts and stops the heating and / or dehumidification device by comparing the difference between the two with the preset threshold value; The temperature and humidity collection device collects on-site environmental temperature and humidity data for the local controller to calculate the dew point temperature, wherein the temperature and humidity collection device is also used to collect the surface temperature of the protected part; The heating device is used to raise the environmental temperature; The dehumidification device is used to reduce the environmental humidity; The communication module is used for data transmission between the local controller and the remote prediction terminal, and sends the on-site data and receives the prediction information; The remote prediction terminal obtains the on-site data through the communication module; The field controller connects to the temperature and humidity acquisition device to obtain ambient temperature and humidity data and calculate the dew point temperature. It also obtains the surface temperature of the protected area. By comparing the difference between the two values ​​with a preset threshold, it determines the condensation risk and then starts and stops the heating and / or dehumidification devices. Local autonomous control is prioritized, and remote prediction information is integrated to optimize decision-making. A programmable logic controller (PLC) can be used. The temperature and humidity acquisition device is used to collect temperature and humidity data of the on-site environment and provide basic data for the on-site controller to calculate the dew point temperature; The heating device is used to increase the ambient temperature when condensation risk occurs to increase the difference between the surface temperature and the dew point temperature. Anti-condensation heaters such as constant temperature heating plates or heating rods can be used; Dehumidification devices are used to reduce ambient humidity and also to improve the above-mentioned temperature difference. Micro dehumidifiers or "fan + desiccant" combination devices can be used; The communication module is used for data transmission between the field controller and the remote prediction terminal. It regularly sends field data to the remote prediction terminal and receives environmental change prediction information from the remote prediction terminal. A 4G cellular network module is preferred, but wired Ethernet, Wi-Fi, etc. can also be used. The remote prediction terminal obtains on-site data through the communication module, uses the built-in LSTM deep learning model to predict the trend of ambient temperature and humidity changes within a predetermined time in the future, assesses the condensation risk, or sends prediction parameters to the on-site controller to provide auxiliary decision-making information. It can be set up in a remote control center such as an industrial computer or cloud server.

[0016] During actual deployment, the field controllers use Nanda Aotuo NA200 series and Siemens 200-smart series PLC programmable logic controllers. Through ladder diagrams or written control logic, data verification, dew point temperature calculation, and threshold comparison function modules are integrated into the program to realize the processing of temperature and humidity data and logical judgment of device start and stop. The temperature and humidity acquisition device uses SHT30 industrial-grade temperature and humidity sensor, which is connected to the PLC via RS485 interface. The hardware is directly installed in the key area inside the protected equipment. The sampling frequency of the sensor is configured in the software to ensure real-time data. The surface temperature sensor uses PT100 probe, which is directly installed near the protected area and connected to the PLC via PT100 interface or acquisition module. The engineering quantity conversion is set in the program to ensure the accurate transmission of surface temperature data. The heating device uses 50-100W anti-condensation heating plate or heating rod, and the dehumidification device uses a micro dehumidifier or "fan + desiccant" combination device. Both are connected to the digital output terminal of the PLC through relays. The software controls the on and off of the relay to realize the start and stop of the device.

[0017] The communication module is preferably a 4G DTU CAT1 / cat4 module such as a person DR512, or a wired Ethernet or Wi-Fi module can also be used according to the site conditions, connected to the PLC through the Ethernet interface, and the MQTT communication protocol is configured in the software to realize the periodic uploading of the site data to the remote prediction terminal, and to receive the remote prediction information, the remote prediction terminal is deployed on the industrial computer or cloud server of the remote control center, the LSTM deep learning model is built through Python, and the data receiving, model training and prediction, risk assessment and prediction information sending functions are realized in the software, and sufficient computing resources are configured on the hardware to support the real-time operation of the model.

[0018] During deployment, first complete hardware installation: fix the temperature and humidity sensor and the surface temperature sensor at the specified position of the protected equipment, connect the communication line between the sensor and the PLC, connect the heating device and the dehumidifying device to the output end of the PLC through the relay, install the communication module and connect it to the PLC and the network; then perform software configuration: write and download the control program in the PLC programming software, set the sampling frequency, threshold parameter, communication period, etc., deploy the deep learning model on the remote terminal server, and configure the network communication parameters to ensure normal data interaction with the on-site controller; finally, perform system debugging, including sensor calibration, PLC logic test, communication link test, to ensure that all parts of the system work together to meet the anti-condensation control requirements.

[0019] Example 2 Further illustrated in combination with Example 1, as shown in Figs. 1-3 S1, obtaining the on-site environment temperature and humidity data, calculating the dew point temperature through the on-site controller, and simultaneously obtaining the surface temperature of the protected part; S2, judging the condensation risk according to the difference between the surface temperature and the dew point temperature, if the difference is lower than the preset safety threshold, then starting the heating device or the dehumidifying device to adjust the environmental conditions through the on-site controller; S3, transmitting the on-site data to the remote prediction terminal through the communication module, and receiving the environmental change prediction information of the remote prediction terminal; S4, predicting the change trend of the environmental temperature and humidity in the future predetermined period of time through the remote prediction terminal based on the deep learning model, and sending the prediction information to the on-site controller; S5, executing the control decision through the on-site controller by comprehensively considering the prediction information and the real-time data, wherein the local real-time control is preferentially executed, and the prediction information is used as an auxiliary optimization decision.

[0020] The method realizes local autonomous control based on real-time data through S1 and S2, ensures timely response to condensation risk in any case, and avoids safety hazards when the remote part fails; S3 and S4 introduce remote deep learning prediction, which can predict the trend of environmental changes in advance, making the control more forward-looking; in S5, the local control priority is combined with remote prediction assistance, which not only ensures the reliability of the system, but also reduces unnecessary start and stop of the heating device and the dehumidifying device, reduces energy consumption and equipment wear and tear, and improves the intelligence level and stability of the condensation prevention control.

[0021] In the preferred scheme, the condensation risk is judged according to the difference between the surface temperature and the dew point temperature in step S2, comprising: The field controller periodically acquires environmental temperature and humidity data; The dew point temperature of the current environment is calculated according to the environmental temperature and humidity data, and the dew point temperature calculation uses the formula , wherein f is the dew point temperature based on the dew point temperature The surface temperature of the protected part is acquired, and the difference between the surface temperature and the dew point temperature is calculated, and the difference formula is: , wherein is the surface temperature , and is the dew point temperature; If the difference is lower than a preset safety threshold, it is determined that there is a condensation risk, and the heating device or the dehumidifying device is triggered to start; During the operation of the heating device or the dehumidifying device, the difference is continuously monitored until the difference is higher than another preset threshold, and the corresponding device is stopped.

[0022] In step S2, the field controller acquires environmental temperature and humidity data at a set period, calculates the dew point temperature of the current environment according to the formula, acquires the surface temperature of the protected part, and calculates the difference between the two by the formula; if the difference is lower than a preset safety threshold, it is determined that there is a condensation risk, and the heating device or the dehumidifying device is triggered to start; during the operation of the device, the difference is continuously monitored until it is higher than another preset threshold, and the corresponding device is stopped.

[0023] By periodically acquiring data and calculating the dew point temperature and the difference, real-time monitoring and judgment of the condensation risk are realized, ensuring that potential risks can be discovered in time; the dew point temperature and the difference are calculated by a clear formula, making the risk judgment more objective and accurate; setting two groups of thresholds for starting and stopping avoids frequent start and stop of the device when the difference fluctuates around the critical value, reduces equipment wear and tear and energy consumption, and improves the stability and reliability of the system operation In the preferred solution, the heating device or the dehumidifying device is started by the on-site controller in step S2 to adjust the environmental conditions, including: According to the source of the condensation risk, the dominant factor of the environmental temperature or humidity is determined; If the condensation risk is mainly caused by low environmental temperature, the heating device is preferentially started by the on-site controller to increase the temperature; If the condensation risk is mainly caused by high environmental humidity, the dehumidifying device is preferentially started by the on-site controller to reduce the humidity; If the condensation risk is affected by both temperature and humidity, the heating device and the dehumidifying device are simultaneously started to quickly adjust the environmental conditions.

[0024] After the on-site controller determines that there is a condensation risk, it first analyzes which is the dominant factor of the environmental temperature or humidity according to the source of the condensation risk. If it is determined that the condensation risk is mainly caused by low environmental temperature, the heating device is preferentially started by the on-site controller to increase the environmental temperature. If it is mainly caused by high environmental humidity, the dehumidifying device is preferentially started by the on-site controller to reduce the environmental humidity. If the condensation risk is affected by both temperature and humidity, the heating device and the dehumidifying device are simultaneously started to quickly adjust the environmental conditions.

[0025] Through the judgment of the dominant factor of the condensation risk, the corresponding heating or dehumidifying device can be started specifically, avoiding energy waste and equipment wear caused by blind starting of the device. Preferential starting of the device corresponding to the dominant factor can improve the efficiency of environmental regulation and quickly alleviate the condensation risk. When both temperature and humidity affect, the two devices are simultaneously started, which can minimize the adjustment time and ensure that the protected part quickly escapes from the condensation risk, improving the accuracy and effectiveness of the condensation prevention control.

[0026] In the preferred solution, the on-site data is transmitted to the remote prediction terminal by the communication module in step S3, including: The on-site environmental temperature, humidity, dew point temperature, surface temperature, and equipment running state data are regularly collected by the communication module; The on-site data is transmitted to the remote prediction terminal for storage and analysis; The prediction information generated by the remote prediction terminal based on the on-site data is received by the communication module; The prediction information is fed back to the on-site controller for subsequent decision assistance; The communication module uses a 4G cellular network module, and the data transmission period is every 1-2 minutes.

[0027] In step S3, the communication module regularly collects data such as on-site ambient temperature, humidity, dew point temperature, surface temperature, and equipment operating status every 1-2 minutes, and transmits this on-site data to the remote prediction terminal, which stores and analyzes it. At the same time, the communication module receives the environmental change prediction information generated by the remote prediction terminal based on the on-site data, and feeds this prediction information back to the on-site controller to assist the on-site controller in its subsequent decision-making.

[0028] By adopting 4G cellular network modules for data transmission, the stability and applicability of communication between on-site and remote terminals are guaranteed, which is especially suitable for the complex network environment of industrial sites. Setting a transmission cycle of every 1-2 minutes can not only update on-site data and forecast information in a timely manner, but also avoid the waste of network resources caused by too frequent transmission. Regularly collecting and transmitting multi-dimensional on-site data provides a comprehensive basis for the analysis and prediction of remote forecast terminals, and feeding back forecast information to the on-site controller makes on-site control decisions more forward-looking, improving the coordination and intelligence level of the system.

[0029] In the preferred solution, in step S4, the change trend of the ambient temperature and humidity within a predetermined time period in the future is predicted by the remote prediction terminal based on the deep learning model, including: Obtain field data as input through remote prediction terminals; Feeding field data into pre-trained deep learning models for time series analysis; Outputting the ambient temperature and humidity change trend data within a predetermined time period in the future through a deep learning model. The deep learning model may adopt an algorithm structure suitable for time series prediction, such as a long short-term memory network (LSTM) or a temporal convolutional network (TCN). The deep learning model is used to output the ambient temperature and humidity change trend data within a predetermined time period in the future, such as predicting the expected temperature every 5 minutes in the next hour. and humidity ; Generate forecast information based on change trend data and send it to the field controller through the communication module; In the preferred solution, in step S5, the field controller integrates the prediction information and real-time data to execute the control decision, including: receiving forecast information via a field controller; Combine forecast information with current real-time data for comprehensive analysis; If the forecast information indicates that the risk of condensation will increase in the future, the heating device or dehumidification device will be started in advance for preventive adjustment; If forecast information indicates that environmental conditions are about to improve, delay the start of heating or dehumidification devices to reduce unnecessary operation; In any case, the safety control logic is executed based on the difference of real-time data first; In the preferred embodiment, the specific steps of steps S4-S5 are: Field data includes ambient temperature in time series form , relative humidity , dew point temperature , surface temperature of key parts And the operating status mark of the heating / dehumidification device After cleaning the data, cubic spline interpolation is used to fill missing values, and then the sliding window statistics, temperature and humidity change rates, and the cumulative amount of the difference between dew point and surface temperature are extracted and spliced ​​into a high-dimensional input vector ; The input vector is input into the pre-trained improved LSTM model (including attention mechanism) for time series analysis. The model input contains The historical data window of time steps is calculated by the attention layer, and the root mean square error is used As the loss function training, after multi-scale residual correction, the output is the ambient temperature forecast value every 5 minutes in the next 1 hour. and humidity forecast values , and estimated 95% confidence interval based on the Bootstrap method; According to the predicted value, the improved Magnus-Tetens formula is used Calculate future dew point temperature and define condensation risk index ,in is the safety threshold, For the danger threshold, prediction information including the predicted value, risk index and confidence interval is generated and sent to the field controller through the communication module; in, is the dew point temperature at each future time step, is the condensation risk index; After receiving the prediction information, the field controller first analyzes the predicted ambient temperature values ​​for each future time step. , humidity prediction value , dew point temperature at each future time step , condensation risk index And confidence interval, calculate the credibility score C of the predicted data: Calculating credibility scores ,in is the variance of the temperature prediction value, is the reference variance, ; Construct a multi-objective optimization model with the goals of "minimizing condensation risk" and "minimizing equipment operating energy consumption" and define decision variables , the objective function ,in is the weight, is the energy consumption function; Solving the optimal strategy based on dynamic programming , combined with real-time temperature difference Perform fusion correction, if The device is forced to start, otherwise it is decided to start in advance or postpone it according to the optimal action and prediction credibility, and the prediction model is corrected online based on real-time data to finally output the device control signal.

[0030] Step S4: Remote prediction terminal predicts temperature and humidity trends based on deep learning model Field data preprocessing and feature engineering: The field data obtained by the remote prediction terminal includes the ambient temperature in the form of time series , relative humidity , dew point temperature , surface temperature of key parts , and the operating status mark of the heating / dehumidification device , where runtime , when stopped , the data sampling interval is 5 minutes. First, the data is cleaned, outliers are removed, and missing values ​​are filled using cubic spline interpolation. The formula is: Let the missing point be , known adjacent points 、 、 、 , then the interpolation function , by solving the system of equations to determine the coefficients , to ensure data continuity. Then extract features, including sliding window statistics, temperature and humidity change rate 、 , and the cumulative difference between the dew point and the surface temperature , concatenate these features with the original data into a high-dimensional input vector .

[0031] Training and optimization of deep learning models: The improved LSTM model is used as the prediction core, and the model input includes time-step historical data window , output for the future The temperature of each time step (e.g., 12 time steps, corresponding to 1 hour) and humidity The model structure includes: input layer dimensions and Consistency, embedding layer, 2-layer LSTM unit, attention layer, calculate the weight of each historical time step , the formula is ,in , For the The LSTM hidden state of time steps, are learnable parameters and fully connected layers (outputting temperature and humidity prediction values). The root mean square error (RMSE) is used as the loss function during training: ,in is the number of samples, and the parameters are iteratively optimized by the Adam optimizer until the loss converges.

[0032] Multi-scale residual correction and confidence interval estimation: To improve the prediction accuracy, a multi-scale residual correction module is added after the LSTM output. First, the residual sequence of the model's initial prediction value and the historical true value is calculated. (The same applies to humidity), and then use wavelet transform to perform multi-scale decomposition of the residual (for example, using db4 wavelet basis, decomposition into 3 layers), and train the autoregressive model (AR(p)) for each scale residual: ,in is the scale index, is a constant term, is the autoregressive coefficient, is white noise. The total residual correction is obtained by superimposing the residual prediction values ​​of each scale The final prediction value is At the same time, the prediction confidence interval is estimated based on the Bootstrap method: the training data is sampled with replacement to generate bootstrap samples, and trained the model separately to obtain Group prediction values, taking the 2.5% and 97.5% quantiles for each time step as the 95% confidence interval , used to assess forecast uncertainty.

[0033] Condensation risk index calculation and forecast information generation: Based on the predicted temperature and humidity, calculate the dew point temperature at each future time step , using the improved Magnus-Tetens formula: ,in , (Correction coefficient for temperatures between 0 and 50°C). Further definition of condensation risk index ,in is a safety threshold (such as 5°C), is the danger threshold (such as 3℃), when hour, (High risk); when hour, (Risk-free). The predicted values ​​for temperature, humidity, dew point, risk index, and confidence interval for the next hour are packaged as forecast information and sent to the field controller via the communication module.

[0034] Step S5: On-site controller synthesizes decision and regulation algorithm Prediction information analysis and credibility evaluation: After receiving the prediction information, the on-site controller first analyzes the future time steps 、 、 、 and confidence intervals, and calculates the credibility score of the prediction data : where is the variance of the temperature prediction value (calculated from the confidence interval: ), is the preset reference variance (such as 2℃²), is the weight coefficient. When (such as 0.3), it is determined that the prediction credibility is low and only serves as a weak reference; otherwise, it is a strong reference participating in decision-making.

[0035] Multi-objective optimization decision model construction: Taking "minimizing condensation risk" and "minimizing equipment operation energy consumption" as the target, the optimization model is constructed. Define the decision variable (0: do not start the equipment, 1: start the heating device, 2: start the dehumidification device, 3: start at the same time), and the objective function is: where is the weight, is the risk index in the step after taking action (based on prediction data correction), is the energy consumption function (such as heating device energy consumption , is the rated power; dehumidification device energy consumption , is the rated power; when starting at the same time ). The constraint conditions include: equipment minimum operation time constraint (if , then , is the minimum operation time), state transition constraint, and minimum downtime interval needs to be met from starting to stopping.

[0036] Optimal action selection based on dynamic programming: dynamic programming is used to solve the above optimization model, and the state variable is defined, where is the current equipment running time. The state transition equation is: , where if starting heating (otherwise 0), if starting dehumidification (otherwise 0), is the heating device temperature rise coefficient, is the environmental heat dissipation coefficient, is the external environment temperature, is the dehumidification coefficient of the dehumidification device, is the humidity exchange coefficient, is the external environment humidity, is the random disturbance.

[0037] value function wherein is the instantaneous cost (combining risk and energy consumption), is the discount factor (such as 0.9). The optimal strategy is solved by backward induction , that is, the current control action to be taken.

[0038] fusion correction of real-time data and prediction: the optimal action obtained by dynamic programming is fused with real-time monitoring data, if the real-time temperature difference , the device is forced to start (priority to meet safety); otherwise, if the optimal action is to start and the prediction credibility , the device is started in advance; if the optimal action is not to start and , the start is suspended. At the same time, the prediction model is online corrected based on real-time data: the deviation , is the prediction advance, the prediction correction coefficient is updated, is the learning rate, used to adjust the weight of the next round of prediction. Finally, the decision result is converted into a device control signal (start / stop), and the decision basis and device state are recorded.

[0039] In the preferred solution, in step S5, the local real-time control is preferentially executed, including: when the remote prediction terminal or communication module fails, the anti-condensation control is independently executed by the on-site controller based on real-time data; the difference between the surface temperature and the dew point temperature is continuously monitored by the on-site controller; if the difference is lower than the preset safety threshold, the heating device or the dehumidification device is immediately started; the local control logic is maintained by the on-site controller to ensure that the system still operates stably when the remote assistance fails.

[0040] In the preferred solution, when the remote prediction terminal or communication module fails, the anti-condensation control is independently executed by the on-site controller based on real-time data, the on-site controller continuously obtains the environmental temperature and humidity data through the temperature and humidity acquisition device, calculates the dew point temperature, and continuously monitors the difference between the surface temperature of the protected part and the dew point temperature through the surface temperature sensor. The field controller has preset start and stop thresholds. When the difference falls below the start threshold, the heating device or dehumidification device is immediately activated. The dominant factor of condensation risk is determined by analyzing the deviation between the ambient temperature and the normal operating temperature range of the equipment, and the deviation between the ambient humidity and the safe humidity range. The corresponding device is activated first. If both temperature and humidity exceed the standard, they are activated simultaneously. After the heating device and dehumidification device are started, the field controller executes the equipment protection logic and sets the minimum continuous operation time. Even if the difference rises above the start threshold during this period, it must run until the minimum continuous operation time ends. When the running time exceeds the minimum continuous operation time, the field controller compares the difference with the stop threshold. If the difference continues to be higher than the stop threshold and remains for the preset time, a stop command is issued; otherwise, the operation continues; The field controller records the time of each start and stop, the operating duration, and the difference change curve data and stores them in the local cache. At the same time, it monitors the operating status of the heating device and the dehumidification device. If a device continuously runs beyond the preset maximum single operating limit or the number of starts and stops exceeds the preset number within the preset time, an early warning will be issued through the local indicator light. After the remote prediction terminal or communication module recovers from the fault, the field controller uploads the local cache data to the remote prediction terminal through the communication module. During the entire process, the field controller maintains the above-mentioned local control logic to ensure that the system still operates stably when remote assistance fails.

[0041] Through the above-mentioned local real-time control priority execution scheme, in the event of a remote partial failure, the on-site controller can operate independently and reliably based on real-time data, ensuring that the anti-condensation function is not interrupted and equipment safety is guaranteed; by pre-setting start and stop thresholds and analyzing risk-dominant factors, the device can be accurately started and stopped, reducing unnecessary energy consumption and equipment wear; settings such as the minimum continuous operation time in the equipment protection logic avoid frequent start and stop of the device, extending its service life; local cache data recording and status monitoring facilitate fault tracing and equipment maintenance, improving system stability and maintainability; data upload after remote recovery ensures data continuity, provides support for remote prediction model optimization, and overall improves the system's fault tolerance and reliability.

[0042] Example 3 Further illustrate with reference to Example 1, Figs. 1-3 As shown, a system and method for controlling temperature and humidity to prevent condensation, combining local autonomous control with remote deep learning prediction-assisted control and decision-making, is provided. In this invention, local field control operates independently as the highest priority link, while remote prediction provides forward-looking decision-making information. If the remote control fails, the field control can still autonomously complete the anti-condensation function, thereby improving the stability and intelligence of the system.

[0043] To achieve the above object, the application provides a condensation-proof temperature and humidity control system, comprising a field controller, a temperature and humidity acquisition device, a heating device, a dehumidification device, a communication module and a remote prediction terminal. The field controller (such as a PLC programmable logic controller) is connected to the temperature and humidity acquisition device to obtain environmental temperature and humidity data and calculate the corresponding dew point temperature; the field controller also obtains the surface temperature of the protected part (such as the high-voltage contact) and compares the surface temperature with the dew point temperature. When the temperature difference between the two is lower than a preset safety threshold, the field controller starts the heating device and / or the dehumidification device operating in the protected environment to raise the environmental temperature or reduce the environmental humidity; the devices will continue to work after being started until the above temperature difference is restored to be higher than another preset threshold, so as to avoid repeated start-stop caused by premature shutdown. In the local control process, the heating device and the dehumidification device can be independently controlled by the field controller: when the main condensation risk comes from the low environmental temperature, the heating device is preferentially started, and when the main risk comes from the high humidity, the dehumidification device is preferentially started; if necessary, both devices can be started at the same time to eliminate the condensation risk at the fastest speed. In addition, the minimum continuous working time and the minimum shutdown interval of the equipment are set in the field control strategy to prevent unnecessary loss caused by frequent start-stop of the equipment.

[0044] The condensation-proof temperature and humidity control system further comprises a remote prediction terminal for predicting the trend of the field environmental change. The remote prediction terminal regularly obtains the data of the temperature, humidity, dew point temperature, surface temperature and condensation-proof device working state of the field through the communication module (preferably a 4G cellular network module to avoid using special protocols). The remote prediction terminal is built-in with a deep learning model for predicting the change trend of the environmental temperature and humidity in a predetermined future time (such as the next 1 hour) based on the collected field historical data. The deep learning model can adopt any algorithm structure suitable for time series prediction (such as long short-term memory network LSTM, time series convolution network TCN, etc.), and is not limited to a specific algorithm. The remote prediction terminal sends the calculated future environmental temperature and humidity prediction results or condensation risk evaluation information back to the field controller through the communication module for decision reference.

[0045] After receiving the prediction information sent by the remote prediction terminal, the field controller will make a comprehensive judgment based on the predicted value and the current real-time monitoring value to optimize the control decision. For example, when the remote prediction shows that the ambient humidity is about to rise and may cause condensation risk, the field controller can start the dehumidification device in advance to actively reduce the humidity; conversely, if the environment is predicted to naturally heat up or become dry, and the current temperature difference is still near the critical threshold, the field controller can temporarily postpone starting the heating or dehumidification device, thereby reducing unnecessary starts and stops. Local autonomous control is always the priority guarantee - regardless of whether remote prediction information is received, the field controller will independently make anti-condensation control actions based on the real-time calculated dew point temperature and the preset threshold; remote prediction information is only used as an auxiliary decision input to make the control process more forward-looking and stable while ensuring safety.

[0046] By adopting an architecture that prioritizes local control and is assisted by remote prediction, the system of the present invention has good fault tolerance and redundant design. Once a remote prediction terminal or communication link fails, the on-site controller can still rely on local sensor data to continue to perform anti-condensation control according to the established threshold strategy, and the safe operation of on-site equipment will not be affected by the failure of the remote part. At the same time, thanks to the introduction of remote deep learning prediction, this system can effectively reduce the number of unnecessary starts and stops of heaters and dehumidifiers, extend the service life of equipment and reduce energy consumption. In summary, the present invention combines the reliability of traditional on-site real-time control with the foresight of artificial intelligence prediction, significantly improving the intelligence level and operational stability of industrial on-site anti-condensation.

[0047] The present invention also provides an anti-condensation temperature and humidity control method, whose steps generally correspond to the control process of the above-mentioned system, including: obtaining on-site ambient temperature and humidity data and calculating the dew point; monitoring the surface temperature of key parts and comparing the temperature difference to decide whether to start or stop the anti-condensation device; reporting on-site data to a remote terminal via a communication module; using a deep learning model to predict future environmental trends; and sending the prediction results back to the site for the controller to refer to for decision-making. The on-site controller adjusts the control strategy based on the real-time data and prediction information. When remote prediction information is unavailable, it degenerates to local control based solely on real-time thresholds. This method fully embodies the concept of combining local autonomous control with remote intelligent prediction, optimizing the anti-condensation control effect while ensuring safety.

[0048] Example 4 Further illustrate with reference to Example 1, Figs. 1-3 As shown, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments are only used to explain part of the present invention and are not intended to limit the scope of protection of the present invention. Various modifications or equivalent substitutions made by those skilled in the art under the guidance of the present invention without departing from the principles of the present invention should be deemed to fall within the scope of protection of the present invention.

[0049] As shown in Fig. 1 The anti-condensation temperature and humidity control system provided by the embodiment includes a field controller 1, an environment temperature and humidity sensor 2, a key position surface temperature sensor 3, a heating device 4, a dehumidification device 5, a communication module 6 and a remote prediction terminal 7. The field controller 1 preferably adopts a PLC (Programmable Logic Controller), but other embedded devices with logic operation and control functions can also be used. The environment temperature and humidity sensor 2 is installed in the internal space of the protected equipment to detect the temperature (Ta) and relative humidity (RH) of the environment in real time.

[0050] The key position surface temperature sensor 3 is installed at the key component (such as the surface of the high-voltage contact) that needs to be prevented from condensation to detect the current surface temperature (Ts) of the component. The heating device 4 can adopt an anti-condensation heater (such as a constant temperature heating sheet or a heating rod) to raise the local temperature inside the equipment; the dehumidification device 5 can adopt a miniature dehumidifier or a combination of a "fan + desiccant" device to reduce the humidity of the environment.

[0051] The communication module 6 is used for data communication between the field controller and the remote prediction terminal, and preferably uses a cellular wireless communication module (such as a 4G module), and of course wired Ethernet, Wi-Fi, etc. can also be used to realize communication connection when conditions permit. The remote prediction terminal 7 can be set in a remote control center and has the functions of storing field data and running deep learning prediction algorithms to provide environment trend prediction information.

[0052] The field controller 1 has a program for anti-condensation control stored in advance, and its control flow is as shown in Fig. 2 Firstly, the field controller periodically reads the current environment temperature Ta and relative humidity RH from the environment temperature and humidity sensor 2, and calculates the dew point temperature Td of the current environment according to the two parameters (the calculation of the dew point temperature can adopt existing mature algorithms, such as the approximate formula based on temperature and humidity). Then, the field controller obtains the actual surface temperature Ts of the protected component from the surface temperature sensor 3, and calculates the difference ΔT = Ts - Td between the surface temperature and the dew point temperature. Then, the controller compares ΔT with the preset first threshold value (safety threshold value).

[0053] If ΔT is less than the safety threshold value (indicating that the surface temperature of the key component has approached the dew point temperature, and the risk of condensation increases), the controller triggers the anti-condensation measures: starts the heating device 4 and / or the dehumidification device 5 according to the needs. Starting the heating device can raise the temperature of the surface of the protected component and the surrounding air, and starting the dehumidification device can reduce the humidity in the air, thereby increasing the temperature difference ΔT between the surface temperature and the dew point temperature.

[0054] Preferably, the start condition is considered to be met when ΔT is below, for example, 3°C, and the controller should immediately start at least one anti-condensation device; the device will continue to run after it is started until it is allowed to stop when ΔT returns to be greater than a second threshold value (for example, 5°C). By setting a pair of start / stop threshold values such as 3°C and 5°C, a hysteresis control interval can be formed to avoid frequent start-stop of the equipment due to small fluctuations in the critical value of the temperature difference. Of course, the above-mentioned threshold values can be adjusted according to actual needs in different application environments.

[0055] In the present embodiment, the field controller decouples and independently controls the heating device 4 and the dehumidification device 5: when it is detected that the condensation risk mainly comes from the low ambient temperature, the heating device 4 is preferentially started; when the main risk comes from the high ambient humidity, the dehumidification device 5 is preferentially started; if both temperature and humidity factors can cause condensation risk, both devices can be run at the same time to maximize the possibility of reducing the occurrence of condensation. In order to protect the equipment and improve the control effect, the anti-condensation control program also sets a protection period for each start-stop action: each time the heating or dehumidification device is started, it is required to work continuously for at least a predetermined minimum time (for example, at least 5 minutes of operation), even if the conditions improve, it is not closed prematurely; similarly, after each stop, the equipment is also required to be at least stopped for a minimum time, to prevent too frequent start-stop from damaging the equipment or reducing the control effect.

[0056] The function flow of the remote prediction terminal 7 is shown in Fig. 3 The remote terminal keeps data connection with the field controller 1 through the communication module 6, and acquires the latest field environment parameters and equipment state information according to a predetermined period (for example, every 5 minutes, or other time intervals as needed), including the ambient temperature Ta, the ambient humidity RH, the dew point temperature Td calculated by the field controller, the surface temperature Ts of the key components, and the current running state of the heating device 4 and the dehumidification device 5, etc. The remote terminal inputs the above time series data into its internal deep learning prediction model, which can output the prediction results of the ambient temperature and humidity in a future period of time after offline training. In the present embodiment, the prediction model takes 1 hour in the future as the prediction time span, and outputs the change trend or specific prediction values (for example, the expected temperature and humidity values every 5 minutes) of the ambient temperature and humidity in the next 60 minutes. From these prediction data, the change trend of the dew point temperature over time can be further calculated. The remote prediction terminal evaluates the condensation risk degree according to the model output results, or directly sends the predicted future environment parameters back to the field controller 1. It should be noted that in order to reduce the system complexity and ensure real-time performance, the remote terminal does not directly issue control instructions, but sends prediction information for the field controller to refer to decision-making.

[0057] After receiving the environmental prediction information sent by the remote terminal 7, the field controller 1 will take the prediction information into account together with the current real-time measurement data for the condensation prevention and control decision. Specifically, when making a condensation risk judgment based on the current ΔT, the controller will refer to the prediction parameters as an auxiliary basis. For example, if the currently measured ΔT is slightly higher than the safety threshold and has not triggered the condensation prevention device, but the remote prediction prompts that the ΔT may drop below the threshold in the near future, the field controller can choose to start the corresponding device for preventive intervention in advance to avoid possible condensation later; on the contrary, if the current ΔT is close to the threshold trigger point, but the prediction shows that the environmental conditions will improve soon (temperature rises or humidity decreases), the controller can appropriately delay action and temporarily suspend the start of the heating or dehumidification device, thereby reducing the number of unnecessary work of the equipment. In any case, the safety control logic of local real-time monitoring is always given priority to execute - once the real-time detected ΔT has fallen below the safety line (i.e. the condensation risk has indeed existed), the field controller will immediately start the condensation prevention device and will never ignore the current risk because the prediction result shows that the environment may improve later. The role of remote prediction information is to provide more comprehensive background information to make the control strategy more smooth and intelligent, but it does not weaken the immediate response ability of local safety risk. By combining local threshold control with remote prediction regulation, the system reduces unnecessary energy consumption and equipment load as much as possible while ensuring the reliability of condensation prevention.

[0058] In summary, by combining field autonomous control with remote deep learning prediction, the present application realizes efficient prevention and control of condensation risk in industrial field. The field control part is fast in response and reliable in operation, the remote prediction part is intelligent and forward-looking, and the two parts complement each other, which significantly improves the reliability and energy utilization efficiency of the condensation prevention system. In practical application, users can adjust and optimize the type of deep learning model, the setting of threshold parameters, communication frequency, etc. according to specific needs, but these modifications do not exceed the protection scope of the present application. Using the system and method proposed by the present application, the condensation protection level of key electrical equipment in harsh environments can be greatly improved, and the safety and stability of industrial production processes can be ensured.

Claims

1. A remote deep learning prediction-assisted decision-making anti-condensation temperature and humidity control system, characterized by: It includes a field controller, a temperature and humidity acquisition device, a heating device, a dehumidification device, a communication module and a remote prediction terminal; The field controller connects to the temperature and humidity acquisition device to obtain the ambient temperature and humidity and calculate the dew point temperature. It also obtains the surface temperature of the protected area and starts and stops the heating and / or dehumidification device by comparing the difference between the two with the preset threshold. The temperature and humidity acquisition device collects the on-site ambient temperature and humidity data for the on-site controller to calculate the dew point temperature. The temperature and humidity acquisition device is also used to collect the surface temperature of the protected area. The heating device is used to increase the ambient temperature; Dehumidification device is used to reduce the ambient humidity; The communication module is used for data transmission between the field controller and the remote prediction terminal, sending field data and receiving prediction information; The remote prediction terminal obtains on-site data through the communication module.

2. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 1 is characterized by: The method includes: S1. Obtain the on-site ambient temperature and humidity data, calculate the dew point temperature through the on-site controller, and obtain the surface temperature of the protected area at the same time; S2. Determine the condensation risk based on the difference between the surface temperature and the dew point temperature. If the difference is lower than a preset safety threshold, activate a heating device or a dehumidifying device to adjust the environmental conditions through the field controller. S3. Transmitting the on-site data to the remote prediction terminal through the communication module, and receiving the environmental change prediction information from the remote prediction terminal; S4. Predicting the changing trend of ambient temperature and humidity within a predetermined time period in the future based on a deep learning model through a remote prediction terminal, and sending the prediction information to the field controller; S5. The on-site controller integrates the predicted information and real-time data to execute control decisions, wherein local real-time control is executed first and the predicted information is used as an auxiliary optimization decision.

3. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 2 is characterized by: In step S2, the condensation risk is determined based on the difference between the surface temperature and the dew point temperature, including: Periodically obtain ambient temperature and humidity data through the field controller; Calculate the dew point temperature of the current environment based on the ambient temperature and humidity data. The dew point temperature is calculated using the formula ,in f Based on dew point temperature Dew point calculation function with relative humidity RH; Obtain the surface temperature of the protected area and calculate the difference between the surface temperature and the dew point temperature. The difference formula is: ,in is the surface temperature, is the dew point temperature; If the difference If the temperature is lower than the preset safety threshold, it is determined that there is a risk of condensation, triggering the heating device or dehumidification device to start; Continuous monitoring of the difference during operation of the heating or dehumidifying device , until the difference When it is higher than another preset threshold, the corresponding device is stopped.

4. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 2 is characterized by: In step S2, the heating device or the dehumidifying device is activated by the field controller to adjust the environmental conditions, including: Determine the dominant factor of ambient temperature or humidity based on the source of condensation risk; If the condensation risk mainly comes from low ambient temperature, the heating device should be activated by the on-site controller to increase the temperature first; If the condensation risk is mainly caused by high ambient humidity, the dehumidification device should be activated by the on-site controller to reduce the humidity first; If the risk of condensation is affected by both temperature and humidity, activate both the heating and dehumidification devices to quickly adjust the environmental conditions.

5. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 2 is characterized by: In step S3, the on-site data is transmitted to the remote prediction terminal via the communication module, including: Regularly collect on-site ambient temperature, humidity, dew point temperature, surface temperature and equipment operating status data through the communication module; Transmitting on-site data to remote prediction terminals for storage and analysis; receiving, through a communication module, environmental change prediction information generated by a remote prediction terminal based on on-site data; Feedback prediction information to the on-site controller for subsequent decision-making assistance; The communication module uses a 4G cellular network module, and the data transmission cycle is every 1-2 minutes.

6. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 2, characterized in that: In step S4, the remote prediction terminal predicts the changing trend of ambient temperature and humidity within a predetermined time period in the future based on the deep learning model, including: Obtain field data as input through remote prediction terminals; Feeding field data into pre-trained deep learning models for time series analysis; Outputting the ambient temperature and humidity change trend data within a predetermined time period in the future through a deep learning model. The deep learning model may adopt an algorithm structure suitable for time series prediction, such as a long short-term memory network (LSTM) or a temporal convolutional network (TCN). The deep learning model is used to output the ambient temperature and humidity change trend data within a predetermined time period in the future, such as predicting the expected temperature every 5 minutes in the next hour. and humidity ; Generate forecast information based on the changing trend data and send it to the field controller through the communication module.

7. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 6, characterized in that: In step S5, the field controller integrates the prediction information and real-time data to execute control decisions, including: receiving forecast information via a field controller; Combine forecast information with current real-time data for comprehensive analysis; If the forecast information indicates that the risk of condensation will increase in the future, the heating device or dehumidification device will be started in advance for preventive adjustment; If forecast information indicates that environmental conditions are about to improve, delay the start of heating or dehumidification devices to reduce unnecessary operation; In any case, the safety control logic is executed based on the difference in real-time data first.

8. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 7, characterized in that: The specific steps of steps S4-S5 are: Field data includes ambient temperature in time series form , relative humidity , dew point temperature , surface temperature of key parts And the operating status mark of the heating / dehumidification device After cleaning the data, cubic spline interpolation is used to fill missing values, and then the sliding window statistics, temperature and humidity change rates, and the cumulative amount of the difference between dew point and surface temperature are extracted and spliced ​​into a high-dimensional input vector ; The input vector is input into the pre-trained improved LSTM model (including attention mechanism) for time series analysis. The model input contains The historical data window of time steps is calculated by the attention layer, and the root mean square error is used As the loss function training, after multi-scale residual correction, the output is the ambient temperature forecast value every 5 minutes in the next 1 hour. and humidity forecast values , and estimated 95% confidence interval based on the Bootstrap method; According to the predicted value, the improved Magnus-Tetens formula is used Calculate future dew point temperature and define condensation risk index ,in is the safety threshold, For the danger threshold, prediction information including the predicted value, risk index and confidence interval is generated and sent to the field controller through the communication module; in, is the dew point temperature at each future time step, is the condensation risk index; After receiving the prediction information, the field controller first analyzes the predicted ambient temperature values ​​for each future time step. , humidity prediction value , dew point temperature at each future time step , condensation risk index And confidence interval, calculate the credibility score C of the predicted data: Calculating credibility scores ,in is the variance of the temperature prediction value, is the reference variance, ; Construct a multi-objective optimization model with the goals of "minimizing condensation risk" and "minimizing equipment operating energy consumption" and define decision variables , the objective function ,in is the weight, is the energy consumption function; Solving the optimal strategy based on dynamic programming , combined with real-time temperature difference Perform fusion correction, if The device is forced to start, otherwise it is decided to start in advance or postpone it according to the optimal action and prediction credibility, and the prediction model is corrected online based on real-time data to finally output the device control signal.

9. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 7, characterized in that: In step S5, local real-time control is performed first, including: When a remote prediction terminal or communication module fails, the field controller independently performs anti-condensation control based on real-time data; The difference between the surface temperature and the dew point temperature is continuously monitored by the field controller; If the difference is lower than the preset safety threshold, the heating device or dehumidification device is immediately activated; Maintaining local control logic through field controllers ensures stable system operation even when remote assistance fails.

10. The control method of the anti-condensation temperature and humidity control system with remote deep learning prediction and auxiliary decision-making according to claim 9, characterized in that: When a remote prediction terminal or communication module fails, the field controller independently performs anti-condensation control based on real-time data. The field controller continuously obtains ambient temperature and humidity data through the temperature and humidity acquisition device, calculates the dew point temperature, and simultaneously obtains the surface temperature of the protected area through the surface temperature sensor, and continuously monitors the difference between the surface temperature and the dew point temperature. The field controller has preset start and stop thresholds. When the difference falls below the start threshold, the heating device or dehumidification device is immediately activated. The dominant factor of condensation risk is determined by analyzing the deviation between the ambient temperature and the normal operating temperature range of the equipment, and the deviation between the ambient humidity and the safe humidity range. The corresponding device is activated first. If both temperature and humidity exceed the standard, they are activated simultaneously. After the heating device and dehumidification device are started, the field controller executes the equipment protection logic and sets the minimum continuous operation time. Even if the difference rises above the start threshold during this period, it must run until the minimum continuous operation time ends. When the running time exceeds the minimum continuous operation time, the field controller compares the difference with the stop threshold. If the difference continues to be higher than the stop threshold and remains for the preset time, a stop command is issued; otherwise, the operation continues; The field controller records the time of each start and stop, the operating duration, and the difference change curve data and stores them in the local cache. At the same time, it monitors the operating status of the heating device and the dehumidification device. If a device continuously runs beyond the preset maximum single operating limit or the number of starts and stops exceeds the preset number within the preset time, an early warning will be issued through the local indicator light. After the remote prediction terminal or communication module recovers from the fault, the field controller uploads the local cache data to the remote prediction terminal through the communication module. During the entire process, the field controller maintains the above-mentioned local control logic to ensure that the system still operates stably when remote assistance fails.

Images