Ammonia still multi-parameter intelligent monitoring system and method based on Internet of Things

Abstract

The invention relates to the technical field of ammonia still intelligent control, and discloses an ammonia still multi-parameter intelligent monitoring system based on Internet of Things, which is characterized by comprising a sensor sensing module, a data acquisition module, an edge calculation module, a cloud platform module and an execution control module, the data acquisition module is connected with the edge calculation module, the edge calculation module is respectively connected with the cloud platform module and the execution control module, the execution control module is further connected with the sensor sensing module, and the cloud platform module is further connected with an external factory energy management system. According to the ammonia still multi-parameter intelligent monitoring system and the ammonia still multi-parameter intelligent monitoring method based on the Internet of Things provided by the invention, the intelligent monitoring system covering the whole running process of the ammonia still is constructed through real-time sensing of multi-dimensional parameters of the whole tower, deep learning model prediction and self-adaptive optimization control; the technical problems that in the prior art, parameter monitoring is not comprehensive, dynamic regulation and control are lagged, and energy consumption optimization is insufficient are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for ammonia stripping towers, and in particular to an intelligent multi-parameter monitoring system and method for ammonia stripping towers based on the Internet of Things. Background Technology

[0002] As a core piece of equipment for ammonia-nitrogen separation in chemical, environmental protection, and other fields, the precise control of ammonia stripping towers directly affects ammonia recovery rates, energy consumption levels, and environmental indicators. In current industrial production, traditional ammonia stripping tower monitoring systems face multiple technical bottlenecks: Firstly, parameter acquisition has significant limitations, monitoring only local points such as bottom temperature and steam flow, making it difficult to fully present the temperature gradient distribution and mass transfer state of the gas-liquid two-phase flow within the tower, resulting in operators being unable to comprehensively grasp changes in tower operating conditions. Secondly, control strategies lack intelligence; the commonly used PID control method lags behind in responding to dynamic disturbances such as feed fluctuations and tower pressure changes, especially during production load switching, making it difficult to quickly establish a new thermal equilibrium state, often leading to temperature field imbalances and energy waste. Furthermore, existing systems lack the ability to predict complex nonlinear relationships within the tower, unable to anticipate operating trends based on historical operating data, and can only rely on manual experience for post-event adjustments, resulting in unstable ammonia recovery rates and fluctuating wastewater treatment effects, failing to meet the stringent requirements of current industrial production for high efficiency, energy saving, and environmental protection. Summary of the Invention

[0003] This invention provides an IoT-based intelligent monitoring system and method for multi-parameter ammonia stripping towers. By real-time sensing of multi-dimensional parameters across the entire tower, prediction using deep learning models, and adaptive optimization control, an intelligent monitoring system covering the entire operation process of the ammonia stripping tower is constructed, effectively solving the technical problems of incomplete parameter monitoring, lagging dynamic control, and insufficient energy consumption optimization in existing technologies.

[0004] This invention provides an IoT-based intelligent monitoring system for ammonia stripping tower with multiple parameters, including a sensor sensing module, a data acquisition module, an edge computing module, a cloud platform module, and an execution control module. The sensor sensing module is connected to the data acquisition module, the data acquisition module is connected to the edge computing module, the edge computing module is connected to the cloud platform module and the execution control module, the execution control module is also connected to the sensor sensing module, and the cloud platform module is also connected to an external factory energy management system.

[0005] The sensor sensing module includes multiple sensors deployed on the ammonia stripping tower body, used to sense the tower plate temperature, pressure, and flow data in real time, and transmit the data to the data acquisition module via RS485 bus;

[0006] The edge computing module is used to denoise, convert and store the collected data, and to generate audible and visual alarms based on the built-in tower pressure over-limit threshold, and generate control commands for the execution control module.

[0007] The cloud platform module includes a data middleware and an application server. The data middleware is used to store historical data of the tray temperature of the ammonia stripping tower and uses an LSTM neural network model to predict the correlation trend between the bottom temperature and the steam flow rate. The application server is used to deploy a three-dimensional configuration interface to dynamically render the tray temperature cloud map and transmit abnormal tray temperature data to mobile terminals.

[0008] The execution control module is used to control the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the ammonia stripping tower according to the PID control instructions of the edge computing module, and to feed back the status of the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve to the sensor sensing module.

[0009] Furthermore, the sensor sensing module includes a temperature sensor, a pressure sensor, and an electromagnetic flow meter to sense the tray temperature, pressure, and flow rate data of the ammonia stripping tower in real time; the accuracy of the temperature sensor is set to ±0.5 degrees Celsius, and one temperature sensor is deployed on each of the 30 trays of the ammonia stripping tower; the accuracy of the pressure sensor is set to ±0.2%FS, and one pressure sensor is deployed at the top and bottom of the ammonia stripping tower; the accuracy of the electromagnetic flow meter is set to ±0.2%, and one electromagnetic flow meter is deployed at the inlet and reflux outlet of the ammonia stripping tower.

[0010] Furthermore, the edge computing module performs data noise reduction processing on the collected data using the 3σ principle and performs data conversion processing using the Modbus→MQTT protocol; it sets up a loop to store historical collected data for a set time period, and sets up an audible and visual alarm when the tower pressure exceeds the over-limit threshold.

[0011] The edge computing module connects to the cloud platform module via a 5G private network, and simultaneously controls the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the execution control module via hardwiring.

[0012] Furthermore, the edge computing module generates control instructions for the execution control module based on the tray temperature using an MPC algorithm, specifically including:

[0013] Using the MPC algorithm, a PID control command for the steam regulating valve is generated based on the target value set for the 15th tray, so as to control the overshoot to be less than the first set value.

[0014] When the pressure fluctuation inside the ammonia stripping tower is detected to be greater than the second set value, a PID control command for the reflux pump frequency converter is generated to automatically optimize the reflux ratio to stabilize the mass transfer efficiency; wherein, the adjustment range of the reflux ratio is 3-5.

[0015] Furthermore, the cloud platform module employs an LSTM neural network model to predict the correlation trend between the bottom temperature of the tower and the steam flow rate, specifically including:

[0016] The bottom temperature, steam flow rate, operating pressure inside the tower, and feed dilute ammonia water flow rate and temperature of the ammonia stripping tower are collected according to a preset sampling period to form raw feature data and establish an operating condition dataset; wherein, the operating condition dataset includes full load, medium load and low load;

[0017] The original feature dataset is processed by anomaly removal, missing value imputation, data standardization, and derivative feature data construction.

[0018] Based on the LSTM neural network model, we set the input data, key optimizations, loss function and optimizer, and performed model training and dynamic validation on the LSTM neural network model.

[0019] Based on the trained LSTM neural network model, the system takes historical data from the first set time as input, predicts the trend of the bottom temperature of the tower at the second set time in the future, calculates the deviation ΔT between the predicted temperature and the target value, and generates the steam flow rate adjustment ΔQ = k × ΔT × f, where k is the temperature-flow coefficient, f is the tower pressure, and the pressure correction function. Finally, it outputs the steam regulating valve opening suggestion.

[0020] Every third set time interval, the predicted temperature is adjusted by 0.7 × predicted temperature + 0.3 × measured temperature. When the deviation between the measured temperature at the bottom of the ammonia stripping tower and the predicted temperature is greater than 2℃, the incremental learning of the LSTM neural network model is triggered, that is, the latest set amount of data is loaded for fine-tuning.

[0021] Furthermore, the steps of performing anomaly removal, missing value imputation, data standardization, and derived feature construction on the original feature dataset include:

[0022] The 3σ principle was used to remove outliers with bottom temperature >145℃ or <130℃ and steam flow fluctuation >20%, and missing values ​​were filled by forward difference method combined with linear interpolation.

[0023] The original feature data is Min-Max standardized to map to the [0,1] interval, and derived features are constructed, including steam flow rate change rate, tower pressure-temperature coupling coefficient, and feed-steam hysteresis factor:

[0024] The steam flow rate change rate is set to (current flow rate - previous time step flow rate) / previous time step flow rate;

[0025] The tower pressure-temperature coupling coefficient is set to tower pressure × temperature / 100;

[0026] Feed-steam lag factor, set as the time difference between changes in feed flow rate and the response of bottom temperature, in minutes.

[0027] Furthermore, the steps of setting input data, key optimizations, loss functions, and optimizers based on the LSTM neural network model, and training and dynamically validating the LSTM neural network model include:

[0028] Input data: 12 time steps × 8 features, the 8 features include 5 original feature data + 3 derived feature data;

[0029] Key optimizations: BatchNormalization alleviates gradient vanishing, and Dropout suppresses overfitting;

[0030] Loss function: Mixed loss function = 0.7×MSE + 0.3×MAE, to balance the penalty for large errors and small errors;

[0031] Optimizer: Adam is used with a learning rate of 0.001. The momentum parameter is adaptively adjusted and cosine annealing is used to decay the learning rate.

[0032] The training set, validation set, and test set are divided according to a set ratio. Mini-batch training is used, and the total number of iterations is set. Training is terminated when the loss of the validation set does not decrease for 10 consecutive iterations, and the optimal weights are saved.

[0033] Transfer learning is used for different working condition datasets, that is, pre-training on the full-load dataset first, and then fine-tuning the training on the medium-load and low-load datasets.

[0034] This invention provides an IoT-based intelligent monitoring method for multiple parameters of an ammonia stripping tower. Based on the IoT-based intelligent monitoring system for multiple parameters of an ammonia stripping tower described above, the method specifically includes:

[0035] S1. The sensor sensing module senses the temperature, pressure, and flow rate data of the ammonia stripping tower trays according to a set cycle, and transmits the data to the data acquisition module via RS485 bus;

[0036] S2. The data acquisition module transmits the data to the edge computing module. The edge computing module performs noise reduction, conversion and storage on the acquired data, and generates an audible and visual alarm based on the built-in tower pressure over-limit threshold, and generates control instructions for the execution control module.

[0037] S3. The edge computing module performs audible and visual alarms based on the built-in tower pressure over-limit threshold, and provides temperature warnings based on the deviation rate between the current operating conditions and the optimal temperature gradient.

[0038] S4. The edge computing module generates control commands for the steam regulating valve and the reflux pump frequency converter based on the tray temperature using the MPC algorithm, and generates control commands for the tower top pressure relief valve based on the tower pressure, so as to realize safety emergency linkage.

[0039] S5. The edge computing module transmits data to the cloud platform module. The cloud platform module uses an LSTM neural network model to predict the correlation trend between the bottom temperature of the tower and the steam flow rate, and outputs a steam regulating valve opening suggestion.

[0040] Further, in step S4, generating a control command for the tower top pressure relief valve based on the tower pressure includes:

[0041] When the tower pressure is greater than 0.35 MPa and continues for the set duration, the tower top pressure relief valve will be fully opened automatically, the tower bottom steam valve will be closed in conjunction with the tower pressure relief valve, an emergency shutdown signal will be sent to the plant DCS, and key parameters inside the tower will be uploaded to the external safety monitoring platform via 5G.

[0042] The beneficial effects of this invention are as follows:

[0043] This invention constructs a multi-parameter intelligent monitoring system covering the entire ammonia stripping tower, achieving real-time perception of everything from tray temperature gradients to operating pressure. Combined with LSTM neural networks for nonlinear correlation prediction of bottom temperature and steam flow, and a rolling optimization strategy based on model predictive control, it significantly improves the stability and economy of the ammonia stripping tower operation. Through the collaboration of edge computing and a cloud platform, the system not only achieves responsive and adaptive control of the tower's operating conditions but also learns optimal operating modes under different loads based on historical data, forming a closed-loop management system of data acquisition, intelligent analysis, precise control, and energy efficiency optimization. It effectively solves the problems of incomplete parameter coverage and lagging control in traditional monitoring, reduces production fluctuations through intelligent prediction, and lowers steam consumption through energy optimization algorithms, enabling the ammonia stripping tower to operate efficiently, energy-savingly, and safely, thus improving the intelligence level and environmental adaptability of industrial production. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the structure of the IoT-based intelligent monitoring system for ammonia stripping towers according to the present invention.

[0045] Figure 2 This is a flowchart illustrating the multi-parameter intelligent monitoring method for ammonia stripping tower based on the Internet of Things according to the present invention.

[0046] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0047] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0048] like Figure 1 As shown, the present invention provides an intelligent monitoring system for multiple parameters of an ammonia stripping tower based on the Internet of Things, including a sensor sensing module, a data acquisition module, an edge computing module, a cloud platform module, and an execution control module. The sensor sensing module is connected to the data acquisition module, the data acquisition module is connected to the edge computing module, the edge computing module is connected to the cloud platform module and the execution control module, the execution control module is also connected to the sensor sensing module, and the cloud platform module is also connected to an external factory energy management system.

[0049] (1) Sensor Sensing Module

[0050] The sensor sensing module includes multiple sensors deployed on the ammonia stripping tower body, used to sense the tower plate temperature, pressure, and flow data in real time, and transmit the data to the data acquisition module via RS485 bus.

[0051] The sensor sensing module includes a temperature sensor, a pressure sensor, and an electromagnetic flow meter to sense the tray temperature, pressure, and flow rate data of the ammonia stripping tower in real time.

[0052] Temperature sensor: The accuracy is set to ±0.5 degrees Celsius, and one of the temperature sensors is deployed on each of the 30 trays of the ammonia stripping tower.

[0053] Pressure sensor: The accuracy is set to ±0.2%FS, and one pressure sensor is deployed at the top and one at the bottom of the ammonia stripping tower.

[0054] Electromagnetic flow meter: with an accuracy set to ±0.2%, one of the electromagnetic flow meters is deployed at the inlet and reflux outlet of the ammonia stripping tower.

[0055] (2) Edge computing module

[0056] The edge computing module integrates an edge computing chip to denoise, convert, and store the collected data, and generates audible and visual alarms based on the built-in tower pressure over-limit threshold, and generates control commands for the execution control module.

[0057] The edge computing module performs data noise reduction processing on the collected data using the 3σ principle and performs data conversion processing using the Modbus→MQTT protocol; it sets up a loop to store historical collected data for a set time (72 hours, which can be adjusted according to specific needs and is not limited here), and sets up an audible and visual alarm when the tower pressure exceeds the over-limit threshold (>0.35Mpa);

[0058] The edge computing module connects to the cloud platform module via a 5G private network, and simultaneously controls the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the execution control module via hardwiring (control cycle ≤ 500ms).

[0059] The edge computing module generates control instructions for the execution control module based on the tray temperature using the MPC algorithm. Specifically, it uses the MPC algorithm to generate PID control instructions for the steam regulating valve based on the target value set for the 15th tray (95℃, which can be adjusted as needed and is not limited here) to control the overshoot to be less than the first set value (5%). At the same time, when the pressure fluctuation inside the ammonia stripping tower is detected to be greater than the second set value (0.03 MPa), it generates PID control instructions for the reflux pump frequency converter to automatically optimize the reflux ratio (adjustment range 3-5) to stabilize the mass transfer efficiency.

[0060] When the temperature gradient of the tray is abnormal (e.g., temperature difference <10℃), the steam waste heat recovery system is activated to use the heat from the condensate to preheat the feed dilute ammonia water (increasing the temperature by 5-8℃) and reduce steam energy consumption.

[0061] (3) Cloud Platform Module

[0062] The cloud platform module includes a data middleware platform and an application server. It interfaces with the factory's energy management system via API to synchronize ammonia stripping energy consumption data. The data middleware platform is used to store historical tray temperature data of the ammonia stripping tower (the storage period is longer than that of the edge computing module, and can be set to ≥1 year), and uses an LSTM neural network model to predict the correlation trend between the bottom temperature and steam flow rate. The application server is used to deploy a 3D configuration interface to dynamically render tray temperature cloud maps and transmit abnormal tray temperature data (such as the temperature of the 15th tray deviating from 95℃±5℃) to mobile terminals.

[0063] An LSTM neural network model is used to predict the correlation between bottom temperature and steam flow rate, specifically including:

[0064] Step 1: Collect the bottom temperature, steam flow rate, operating pressure inside the tower, feed dilute ammonia water flow rate and temperature of the ammonia stripping tower according to the preset sampling period (500ms) to form raw feature data and establish an operating condition dataset; wherein, the operating condition dataset includes full load, medium load and low load;

[0065] The data window is in 1-hour increments (12 time steps, each 5 minutes). Each time step includes the average, maximum, minimum, and variance of the parameters within that period. Historical data is labeled according to production load into three categories: full load (≥80%), medium load (50-80%), and low load (<50%). Sub-data sets are created for each category to ensure that the model learns the temperature-flow mapping relationship under different loads.

[0066] Step 2: Perform anomaly removal, missing value imputation, data standardization, and derivative feature data construction on the original feature dataset;

[0067] 1) Anomaly removal and missing value filling: The 3σ principle is used to remove anomalies with bottom temperature >145℃ or <130℃ and steam flow fluctuation >20%, and missing values ​​are filled by forward difference method combined with linear interpolation; for example, if the bottom temperature of a certain time step is missing, the temperature of the previous time step is taken as the steam flow change rate of that period × temperature response coefficient (0.8℃ / % flow change).

[0068] 2) Data Standardization and Derivative Feature Data Construction: The original feature data is Min-Max standardized to map to the [0,1] interval. Simultaneously, derived features are constructed, including steam flow rate change rate, tower pressure-temperature coupling coefficient, and feed-steam hysteresis factor.

[0069] Steam flow rate change rate: set to (current flow rate - previous time step flow rate) / previous time step flow rate;

[0070] Tower pressure-temperature coupling coefficient: set to tower pressure × temperature / 100;

[0071] Feed-steam lag factor: set as the time difference between changes in feed flow rate and the response of bottom temperature, in minutes.

[0072] Step 3: Based on the LSTM neural network model, set the input data, key optimizations, loss function and optimizer, and perform model training and dynamic validation on the LSTM neural network model;

[0073] Input data: 12 time steps × 8 features, the 8 features include 5 original feature data + 3 derived feature data;

[0074] Key optimizations: BatchNormalization alleviates gradient vanishing, and Dropout suppresses overfitting;

[0075] Loss function: Mixed loss function = 0.7 × MSE (mean squared error) + 0.3 × MAE (mean absolute error) to balance the penalty for large errors and small errors.

[0076] Optimizer: Adam is used with a learning rate of 0.001. The momentum parameter is adaptively adjusted and cosine annealing is used to decay the learning rate (20 rounds).

[0077] The training set, validation set, and test set are divided according to a set ratio (7:2:1). Mini-batch training is used, and the total number of iterations is set. Training is terminated when the loss of the validation set does not decrease for 10 consecutive iterations, and the optimal weights are saved.

[0078] Transfer learning is employed for different operating condition datasets. This involves pre-training on a full-load dataset and then fine-tuning the training on medium-load and low-load datasets to enhance the model's adaptability to different operating conditions. For example, under full-load conditions, the response of steam flow to the bottom temperature of the tower lags by approximately 15 minutes; under low-load conditions, the lag extends to 25 minutes. The model automatically adjusts the time delay parameter based on the operating condition label.

[0079] Step 4: Based on the trained LSTM neural network model, input the historical data of the current first set time (1 hour), predict the bottom temperature trend of the tower at the next second set time (30 minutes, 6 time steps), and calculate the deviation ΔT between the predicted temperature and the target value (140℃), generating the steam flow rate adjustment amount ΔQ; ΔQ=k×ΔT×f(tower pressure), where k is the temperature-flow coefficient (0.8% / ℃), f(tower pressure) is the pressure correction function, and finally output the steam regulating valve opening suggestion (e.g., if the current opening is 50%, it is recommended to increase it to 54%).

[0080] Step 5: Every three set time intervals (5 minutes), correct the predicted temperature. The corrected temperature = 0.7 × predicted temperature + 0.3 × measured temperature. When the deviation between the measured temperature at the bottom of the ammonia stripping tower and the predicted temperature exceeds 2℃, trigger incremental learning of the LSTM neural network model, i.e., load the latest set number of data points (100 data points) for fine-tuning. A Kalman filter algorithm is introduced to fuse real-time sensor data and model predictions to generate the optimal control quantity.

[0081] Load the full-condition data of the past 7 days and reclassify it according to the load; update the model weights using incremental learning (retain 80% of the pre-trained weights and fine-tune 20% of the weights related to the new data); if the prediction accuracy is <90%, trigger full data retraining (takes about 2 hours).

[0082] The prediction results are linked with the MPC algorithm: LSTM predicts the temperature trend for the next 30 minutes, and MPC calculates the optimal steam flow regulation sequence (for the next 10 time steps). Rolling optimization is performed every 5 minutes to correct control variables to address disturbances such as feed fluctuations. The control overshoot is ≤5%, and the temperature is stabilized within the range of 140±3℃.

[0083] (4) Execution control module

[0084] The execution control module is used to control the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the ammonia stripping tower according to the PID control instructions of the edge computing module, and to feed back the status of the steam regulating valve (controlling tower bottom heating), reflux pump frequency converter (adjusting reflux ratio), and tower top pressure relief valve (safety protection) to the sensor sensing module.

[0085] like Figure 2 As shown, the present invention also provides an IoT-based intelligent monitoring method for multiple parameters of an ammonia stripping tower. Based on the IoT-based intelligent monitoring system for multiple parameters of an ammonia stripping tower described above, the method specifically includes:

[0086] S1. The sensor sensing module senses the temperature, pressure, and flow data of the ammonia stripping tower trays at a set period (500ms) and transmits them to the data acquisition module via RS485 bus;

[0087] S2. The data acquisition module transmits the data to the edge computing module. The edge computing module performs noise reduction, conversion and storage on the acquired data, and generates an audible and visual alarm based on the built-in tower pressure over-limit threshold, and generates control instructions for the execution control module.

[0088] S3. The edge computing module performs audible and visual alarms based on the built-in tower pressure over-limit threshold, and provides temperature warnings based on the deviation rate between the current operating conditions and the optimal temperature gradient.

[0089] Based on a preset tray temperature gradient model (e.g., a temperature decrease of 15℃ every 10 trays from bottom to top), the deviation rate between the current operating condition and the optimal model is calculated (an early warning is issued when the deviation rate exceeds 15%). For example, if the bottom temperature is <130℃ for 10 minutes: a message is displayed saying "Insufficient steam, it is recommended to increase by 5%"; an audible and visual alarm is triggered and the steam valve is automatically opened by 10%.

[0090] S4. The edge computing module generates control commands for the steam regulating valve and the reflux pump frequency converter based on the tray temperature using the MPC algorithm, and generates control commands for the tower top pressure relief valve based on the tower pressure, so as to achieve safe emergency linkage.

[0091] The edge computing module generates control instructions for the execution control module based on the tray temperature using the MPC algorithm. Specifically, it uses the MPC algorithm to generate PID control instructions for the steam regulating valve based on the target value set for the 15th tray (95℃, which can be adjusted as needed and is not limited here) to control the overshoot to be less than the first set value (5%). At the same time, when the pressure fluctuation inside the ammonia stripping tower is detected to be greater than the second set value (0.03 MPa), it generates PID control instructions for the reflux pump frequency converter to automatically optimize the reflux ratio (adjustment range 3-5) to stabilize the mass transfer efficiency.

[0092] When the tower pressure is greater than 0.35 MPa and continues for the set duration, the tower top pressure relief valve will be fully opened automatically, the tower bottom steam valve will be closed in conjunction with the tower pressure relief valve, an emergency shutdown signal will be sent to the plant DCS, and key parameters inside the tower will be uploaded to the external safety monitoring platform via 5G.

[0093] S5. The edge computing module transmits data to the cloud platform module. The cloud platform module uses an LSTM neural network model to predict the correlation trend between the bottom temperature of the tower and the steam flow rate, and outputs a steam regulating valve opening suggestion.

[0094] This invention constructs a multi-parameter intelligent monitoring system covering the entire ammonia stripping tower, achieving real-time perception of everything from tray temperature gradients to operating pressure. Combined with LSTM neural networks for nonlinear correlation prediction of bottom temperature and steam flow, and a rolling optimization strategy based on model predictive control, it significantly improves the stability and economy of the ammonia stripping tower operation. Through the collaboration of edge computing and a cloud platform, the system not only achieves responsive and adaptive control of the tower's operating conditions but also learns optimal operating modes under different loads based on historical data, forming a closed-loop management system of data acquisition, intelligent analysis, precise control, and energy efficiency optimization. It effectively solves the problems of incomplete parameter coverage and lagging control in traditional monitoring, reduces production fluctuations through intelligent prediction, and lowers steam consumption through energy optimization algorithms, enabling the ammonia stripping tower to operate efficiently, energy-savingly, and safely, thus improving the intelligence level and environmental adaptability of industrial production.

[0095] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0096] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A multi-parameter intelligent monitoring system for ammonia stripping tower based on the Internet of Things, characterized in that, It includes a sensor sensing module, a data acquisition module, an edge computing module, a cloud platform module, and an execution control module. The sensor sensing module is connected to the data acquisition module, the data acquisition module is connected to the edge computing module, the edge computing module is connected to the cloud platform module and the execution control module, the execution control module is also connected to the sensor sensing module, and the cloud platform module is also connected to an external factory energy management system. The sensor sensing module includes multiple sensors deployed on the ammonia stripping tower body, used to sense the tower plate temperature, pressure, and flow data in real time, and transmit the data to the data acquisition module via RS485 bus; The edge computing module is used to denoise, convert and store the collected data, and to generate audible and visual alarms based on the built-in tower pressure over-limit threshold, and generate control commands for the execution control module. The cloud platform module includes a data middleware and an application server. The data middleware is used to store historical data of the tray temperature of the ammonia stripping tower and uses an LSTM neural network model to predict the correlation trend between the bottom temperature and the steam flow rate. The application server is used to deploy a three-dimensional configuration interface to dynamically render the tray temperature cloud map and transmit abnormal tray temperature data to mobile terminals. The execution control module is used to control the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the ammonia stripping tower according to the PID control instructions of the edge computing module, and to feed back the status of the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve to the sensor sensing module.

2. The IoT-based multi-parameter intelligent monitoring system for ammonia stripping towers according to claim 1, characterized in that, The sensor sensing module includes a temperature sensor, a pressure sensor, and an electromagnetic flow meter to sense the temperature, pressure, and flow rate data of the ammonia stripping tower in real time. The accuracy of the temperature sensor is set to ±0.5 degrees Celsius, and one temperature sensor is deployed on each of the 30 trays of the ammonia stripping tower. The accuracy of the pressure sensor is set to ±0.2%FS, and one pressure sensor is deployed at the top and bottom of the ammonia stripping tower. The accuracy of the electromagnetic flow meter is set to ±0.2%, and one electromagnetic flow meter is deployed at the inlet and reflux outlet of the ammonia stripping tower.

3. The IoT-based multi-parameter intelligent monitoring system for ammonia stripping towers according to claim 1, characterized in that, The edge computing module performs data noise reduction processing on the collected data using the 3σ principle and performs data conversion processing using the Modbus→MQTT protocol; it sets up a loop to store historical collected data for a set time period and sets up an audible and visual alarm when the tower pressure exceeds the over-limit threshold. The edge computing module connects to the cloud platform module via a 5G private network, and simultaneously controls the steam regulating valve, reflux pump frequency converter, and tower top pressure relief valve of the execution control module via hardwiring.

4. The IoT-based multi-parameter intelligent monitoring system for ammonia stripping towers according to claim 3, characterized in that, The edge computing module generates control commands for the execution control module based on the tray temperature using the MPC algorithm, specifically including: Using the MPC algorithm, a PID control command for the steam regulating valve is generated based on the target value set for the 15th tray, so as to control the overshoot to be less than the first set value. When the pressure fluctuation inside the ammonia stripping tower is detected to be greater than the second set value, a PID control command for the reflux pump frequency converter is generated to automatically optimize the reflux ratio to stabilize the mass transfer efficiency; wherein, the adjustment range of the reflux ratio is 3-5.

5. The IoT-based intelligent monitoring system for ammonia stripping towers according to claim 1, characterized in that, The cloud platform module uses an LSTM neural network model to predict the correlation trend between the bottom temperature and steam flow rate, specifically including: The bottom temperature, steam flow rate, operating pressure inside the tower, and feed dilute ammonia water flow rate and temperature of the ammonia stripping tower are collected according to a preset sampling period to form raw feature data and establish an operating condition dataset; wherein, the operating condition dataset includes full load, medium load and low load; The original feature dataset is processed by anomaly removal, missing value imputation, data standardization, and derivative feature data construction. Based on the LSTM neural network model, we set the input data, key optimizations, loss function and optimizer, and then performed model training and dynamic validation on the LSTM neural network model. Based on the trained LSTM neural network model, the system takes historical data from the first set time as input, predicts the trend of the bottom temperature of the tower at the second set time in the future, calculates the deviation ΔT between the predicted temperature and the target value, and generates the steam flow rate adjustment ΔQ = k × ΔT × f, where k is the temperature-flow coefficient, f is the tower pressure, and the pressure correction function. Finally, it outputs the steam regulating valve opening suggestion. Every third set time interval, the predicted temperature is adjusted by 0.7 × predicted temperature + 0.3 × measured temperature. When the deviation between the measured temperature at the bottom of the ammonia stripping tower and the predicted temperature is greater than 2℃, the incremental learning of the LSTM neural network model is triggered, that is, the latest set amount of data is loaded for fine-tuning.

6. The IoT-based intelligent monitoring system for ammonia stripping towers according to claim 5, characterized in that, The steps of performing anomaly removal, missing value imputation, data standardization, and derived feature construction on the original feature dataset include: The 3σ principle was used to remove outliers with bottom temperature >145℃ or <130℃ and steam flow fluctuation >20%, and missing values ​​were filled by forward difference method combined with linear interpolation. The original feature data is Min-Max standardized to map to the [0,1] interval, and derived features are constructed, including steam flow rate change rate, tower pressure-temperature coupling coefficient, and feed-steam hysteresis factor. The steam flow rate change rate is set to (current flow rate - previous time step flow rate) / previous time step flow rate; The tower pressure-temperature coupling coefficient is set to tower pressure × temperature / 100; Feed-steam lag factor, set as the time difference between changes in feed flow rate and the response of bottom temperature, in minutes.

7. The IoT-based multi-parameter intelligent monitoring system for ammonia stripping towers according to claim 5, characterized in that, The steps of setting input data, key optimizations, loss function, and optimizer based on the LSTM neural network model, and training and dynamically validating the LSTM neural network model include: Input data: 12 time steps × 8 features, the 8 features include 5 original feature data + 3 derived feature data; Key optimizations: BatchNormalization alleviates gradient vanishing, and Dropout suppresses overfitting; Loss function: Mixed loss function = 0.7×MSE + 0.3×MAE, to balance the penalty for large errors and small errors; Optimizer: Adam is used with a learning rate of 0.

001. The momentum parameter is adaptively adjusted and cosine annealing is used to decay the learning rate. The training set, validation set, and test set are divided according to a set ratio. Mini-batch training is used, and the total number of iterations is set. Training is terminated when the loss of the validation set does not decrease for 10 consecutive iterations, and the optimal weights are saved. Transfer learning is used for different working condition datasets, that is, pre-training on the full-load dataset first, and then fine-tuning the training on the medium-load and low-load datasets.

8. A method for intelligent monitoring of multiple parameters of an ammonia stripping tower based on the Internet of Things, characterized in that, The method of the IoT-based multi-parameter intelligent monitoring system for ammonia stripping tower according to any one of claims 1-7 specifically includes: S1. The sensor sensing module senses the temperature, pressure, and flow rate data of the ammonia stripping tower trays according to a set cycle, and transmits the data to the data acquisition module via RS485 bus; S2. The data acquisition module transmits the data to the edge computing module. The edge computing module performs noise reduction, conversion and storage on the acquired data, and generates an audible and visual alarm based on the built-in tower pressure over-limit threshold, and generates control instructions for the execution control module. S3. The edge computing module performs audible and visual alarms based on the built-in tower pressure over-limit threshold, and provides temperature warnings based on the deviation rate between the current operating conditions and the optimal temperature gradient. S4. The edge computing module generates control commands for the steam regulating valve and the reflux pump frequency converter based on the tray temperature using the MPC algorithm, and generates control commands for the tower top pressure relief valve based on the tower pressure, so as to realize safety emergency linkage. S5. The edge computing module transmits data to the cloud platform module. The cloud platform module uses an LSTM neural network model to predict the correlation trend between the bottom temperature of the tower and the steam flow rate, and outputs a steam regulating valve opening suggestion.

9. The method for intelligent monitoring of multiple parameters of an ammonia stripping tower based on the Internet of Things according to claim 8, characterized in that, In step S4, the control command for generating the pressure relief valve at the top of the tower based on the tower pressure includes: When the tower pressure is greater than 0.35 MPa and continues for the set duration, the tower top pressure relief valve will be fully opened automatically, the tower bottom steam valve will be closed in conjunction with the tower pressure relief valve, an emergency shutdown signal will be sent to the plant DCS, and key parameters inside the tower will be uploaded to the external safety monitoring platform via 5G.

Images