A method, device and equipment for optimizing process parameters of flue gas CO2 capture device

Abstract

The present invention discloses a method, device and equipment for optimizing the process parameters of a flue gas CO2 capture device. The method includes: predicting the key operating parameter data of offline analysis in the amine liquid system of the CO2 capture device based on the real-time measurement data of the CO2 capture device obtained in real time and a pre-trained gated recurrent neural network model; using the predicted key operating parameter data of offline analysis and the real-time measurement data to automatically control and adjust the CO2 capture device and the basis for operators to adjust the operating parameters, so as to optimize and control the process of the CO2 capture device in real time. The deep learning model of the CO2 capture device is used to predict the key parameters that can be measured in real time and require offline analysis, so as to control the operation state of the whole device to a state close to the optimal state, which not only ensures CO2 capture and product purity, but also reduces the amount of public works and energy consumption, reduces solvent loss, thereby reducing the cost of carbon capture and improving the flexibility and convenience of device operation.

Description

Technical Field

[0001] The present invention relates to the field of computer control and environmental protection technology, and in particular to a method, device and equipment for optimizing process parameters of a flue gas CO2 capture device. Background Art

[0002] The flue gas CO2 capture technology currently in the large-scale demonstration and application stage is the chemical absorption method, with composite organic amines as the common absorbent. Common basic process flows include: Figure 1 As shown, the flue gas is sent to the CO2 capture device through the flue gas pipeline, enters the scrubbing tower, and contacts with the water flowing in the countercurrent for water washing. The flue gas that has been cooled and dusted by water washing is sent to the absorption tower through the induced draft fan. The absorption tower is equipped with an inter-stage cooler to increase the absorption capacity. The rich liquid that has absorbed CO2 is sent to the lean-rich liquid heat exchanger through the rich liquid pump. The rich liquid preheated by the lean-rich liquid heat exchanger desorbs CO2 in the regeneration tower, and the CO2 vapor is partially condensed in the condenser. The CO2 gas and liquid water are separated in the separator, and the CO2 gas is sent for subsequent compression. The lean liquid that has desorbed CO2 is cooled by the lean-rich liquid heat exchanger, and further cooled by the lean liquid cooler and circulated into the absorption tower.

[0003] In the above process, several key operating parameters such as CO2 concentration in flue gas, absorption temperature, CO2 content in circulating lean liquid, regeneration tower pressure, etc. play a decisive role in the energy consumption of the entire CO2 capture device. Parameters such as the outlet temperature of the washing tower, the circulation volume of the washing liquid in the water washing section at the top of the absorption tower and the cooling temperature play a decisive role in the water balance of the entire system. The concentration of amine in the washing liquid, the temperature of the exhaust gas at the top of the absorption tower, the concentration of heat-stable salt in the lean liquid and the rich liquid, etc. have a great influence on the loss rate of amine.

[0004] Flue gas conditions usually fluctuate due to various reasons, such as changes in fuel, changes in combustion conditions, changes in the operating conditions of upstream combustion furnaces or industrial furnaces, etc., which will inevitably cause fluctuations in the operating parameters of the CO2 capture device. Chemical plants usually use a distributed control system (DCS) control method to control production operations. However, DCS control can only provide valve action feedback based on online measured conditions such as temperature, pressure, and flow, and the parameters of the amine liquid system (lean liquid and rich liquid) of the CO2 capture device can usually only be obtained through offline sampling laboratory analysis and testing. It cannot be measured online in real time and transmitted to the DCS system, and cannot be optimized and controlled. This will cause problems such as increased energy consumption of the device, substandard CO2 production affecting downstream operations, increased amine liquid loss, and damage to the system's water balance.

[0005] Therefore, intelligent optimization and control of parameters in the design and operation of CO2 capture devices is extremely important in terms of energy conservation and consumption reduction, and stable device operation. Summary of the invention

[0006] In order to be able to predict the key operating parameters of the CO2 capture device so as to adjust the CO2 capture device so that the operating state is close to the optimal state, thereby ensuring CO2 capture and product purity, further reducing public engineering usage and energy consumption, thereby reducing carbon capture costs, and at the same time enriching process routes and increasing selection space, an optimization method, device and equipment for the process parameters of a flue gas CO2 capture device are provided in an embodiment of the present invention.

[0007] In a first aspect, the present invention provides a method for optimizing process parameters of a flue gas CO2 capture device, which may include:

[0008] Based on the real-time measurement data of the CO2 capture device acquired in real time and the pre-trained gated recurrent neural network model, predict the key operating parameter data of the offline chemical analysis in the amine liquid system in the CO2 capture device;

[0009] The predicted key operating parameter data of the offline test analysis and the real-time measurement data are used as the basis for the automatic control and adjustment of the CO2 capture device and the operator to adjust the operating parameters, so as to optimize the process of the CO2 capture device in real time.

[0010] Optionally, the gated recurrent neural network model is pre-trained according to the following steps:

[0011] Based on the process principle and engineering experience of the CO2 capture device, determine the key operating parameters and performance indicators of the CO2 capture device;

[0012] Acquire a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items;

[0013] Preprocessing the key operating parameter data to remove abnormal data contained therein;

[0014] Performing a correlation analysis on the pre-processed key operating parameter data and the performance index data to determine the correlation coefficients between different key operating parameter items; and determining the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients;

[0015] Constructing a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data;

[0016] The gated recurrent neural network is trained using samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

[0017] Optionally, before acquiring a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items, the following step may also be included:

[0018] Collecting measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis and testing data;

[0019] A full-process simulation model of the CO2 capture device is established based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

[0020] Optionally, the establishing of a full-process simulation model of the CO2 capture device based on the measurement and analysis data includes:

[0021] Based on the process of amine liquid absorbing CO2 in the CO2 capture device, an electrolyte property model of the CO2 capture device is established; wherein the electrolyte property model includes: a phase equilibrium model, an absorption chemical reaction model and a thermal stable salt reaction model;

[0022] Selecting an equipment module for drawing a process simulation flow chart based on the electrolyte physical property model;

[0023] Based on the measurement and analysis data, logistics data, equipment parameters and operating conditions, the drawn full-process simulation model is run until the model converges.

[0024] Optionally, the method may also include: comparing the collected measurement and analysis data with the data simulated by the full-process simulation model, and adjusting the model parameters in the electrolyte physical property model based on the comparison results so that the error between the measurement and analysis data and the simulated data is within a preset range.

[0025] Optionally, the method may further include: using a root mean square error and / or a mean absolute error to verify the gap between the predicted value and the true value output by the trained gated recurrent neural network, so as to adjust the model parameters of the gated recurrent neural network.

[0026] In a second aspect, the present invention provides a method for training a machine learning model, which may include:

[0027] Based on the process principle and engineering experience of the CO2 capture device, determine the key operating parameters and performance indicators of the CO2 capture device;

[0028] Acquire a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items;

[0029] Preprocessing the key operating parameter data to remove abnormal data contained therein;

[0030] Performing a correlation analysis on the pre-processed key operating parameter data and the performance index data to determine the correlation coefficients between different key operating parameter items; and determining the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients;

[0031] Constructing a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data;

[0032] The gated recurrent neural network is trained using samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

[0033] Optionally, before acquiring a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items, the following step may also be included:

[0034] Collecting measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis and testing data;

[0035] A full-process simulation model of the CO2 capture device is established based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

[0036] In a third aspect, the present invention provides a device for optimizing process parameters of a flue gas CO2 capture device, which may include:

[0037] A prediction module, for predicting key operating parameter data of offline chemical analysis in the amine liquid system in the CO2 capture device based on real-time measurement data of the CO2 capture device acquired in real time and a pre-trained gated recurrent neural network model;

[0038] The optimization module is used to automatically control and adjust the CO2 capture device based on the predicted key operating parameter data of offline chemical analysis and the real-time measurement data, and to provide a basis for operators to adjust the operating parameters, so as to optimize the process of the CO2 capture device in real time.

[0039] In a fourth aspect, the present invention provides a training device for a machine learning model, which may include:

[0040] A determination module, used to determine key operating parameter items and performance index items of the CO2 capture device based on the process principle and engineering experience of the CO2 capture device;

[0041] An acquisition module, used to acquire a preset number of key operation parameter data corresponding to the key operation parameter items and performance indicator data corresponding to the performance indicator items;

[0042] A preprocessing module, used for preprocessing the key operating parameter data to remove abnormal data contained therein;

[0043] An analysis module, used to perform a correlation analysis on the pre-processed key operating parameter data and the performance indicator data to determine the correlation coefficients between different key operating parameter items; and determine the key operating parameter items that affect the performance indicators of the CO2 capture device based on the correlation coefficients;

[0044] A construction module, used to construct a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data;

[0045] A training module is used to train the gated recurrent neural network with samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

[0046] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for optimizing the process parameters of the flue gas CO2 capture device as described in the first aspect, or implements the method for training the machine learning model as described in the second aspect.

[0047] In a sixth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method for optimizing the process parameters of the flue gas CO2 capture device as described in the first aspect, or implements the method for training the machine learning model as described in the second aspect.

[0048] The beneficial effects of the above technical solution provided by the present invention include at least:

[0049] The present invention provides a method, device and equipment for optimizing the process parameters of a flue gas CO2 capture device. The method collects and utilizes the on-site instrument measurement and on-site analysis data of the device, combines the process flow simulation model for the CO2 capture device, establishes a deep learning model of the CO2 capture device, and then uses the established deep learning model to predict the utility usage, system water balance, and state of amine liquid of the device, and uses the prediction results of the deep learning model in combination with the DCS control system at the device site to intelligently control the operation of the CO2 capture device in real time. The present invention uses the deep learning model of the CO2 capture device to predict key parameters that can be measured in real time and require offline testing and analysis, thereby controlling the operating state of the entire device to a state close to the optimal state, which not only ensures CO2 capture and product purity, but also reduces the utility usage and energy consumption, reduces solvent loss, thereby reducing carbon capture costs, and improves the flexibility and convenience of device operation.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0051] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0053] Figure 1 This is the process flow diagram of the existing CO2 capture device;

[0054] Figure 2 A flowchart of a method for training a machine learning model provided in an embodiment of the present invention;

[0055] Figure 3A flowchart of a detailed method for training a machine learning model provided in an embodiment of the present invention;

[0056] Figure 4 is a specific flow chart of step S32;

[0057] Figure 5 A schematic diagram of the structure of a training device for a machine learning model provided in an embodiment of the present invention;

[0058] Figure 6 A flow chart of a method for optimizing process parameters of a flue gas CO2 capture device provided in an embodiment of the present invention;

[0059] Figure 7 It is a schematic structural diagram of a device for optimizing process parameters of a flue gas CO2 capture device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0061] Embodiment 1

[0062] In order to be able to predict the amine liquid system data that is difficult to measure online in real time in the CO2 capture device, thereby optimizing the key operating parameters of the control device, so that the CO2 capture device can operate at the lowest energy consumption state, the water balance of the CO2 capture device is easy to control, and the amine liquid loss is minimized. In an embodiment of the present invention, a training method for a machine learning model is provided. The deep learning model constructed by the training method can be established based on the Pytorch deep learning library, referring to Figure 2 As shown, the training method may include the following steps:

[0063] Step S21: Based on the process principle and engineering experience of the CO2 capture device, key operating parameter items and performance index items of the CO2 capture device are determined.

[0064] This step is to determine the key operating parameters and performance indicators of the CO2 capture device, wherein the key operating parameters of the CO2 capture device include operating time, CO2 capture rate, amine solution concentration, circulating amine solution flow rate, CO2 content in circulating amine solution, amine solution inlet temperature of the absorption tower, washing liquid flow rate, amine content in the washing liquid, and heat-stable salt concentration, a total of 9 preliminary variables. The performance indicators of the CO2 capture device include steam consumption, circulating water volume, power consumption, solvent loss rate, system drainage, system water replenishment, and amine solution defoaming time, a total of 7 indicators.

[0065] Step S22, obtaining key operating parameter data corresponding to a preset number of key operating parameter items and performance indicator data corresponding to performance indicator items. This step is to obtain a large amount of group data for building a training sample set for training a neural network model. In this embodiment, due to the limited amount of measured data, a large amount of group data is partly derived from measured data and the other part is process simulation data.

[0066] Step S23: pre-process the key operating parameter data to remove abnormal data contained therein.

[0067] In this step, a large amount of data obtained in the above step S22 is preprocessed. In the specific implementation, the 3σ criterion is used to identify and mark outliers in the data. By calculating the mean and standard deviation of the variables, 3 times the standard deviation is used as the random error interval, and the part exceeding this interval is determined as abnormal data. In this step, the above 8 primary selected variables are normalized using the Min-Max normalization method, and the variable value range is between [0, 1] to eliminate the influence of dimensional differences on model accuracy.

[0068] Step S24: perform correlation analysis on the pre-processed key operating parameter data and performance index data to determine the correlation coefficients between different key operating parameter items; and determine the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients.

[0069] In the specific implementation of this step, the distance correlation (DC) coefficient correlation analysis method can be used to conduct in-depth analysis on the data of the above variables and indicators. Through correlation analysis, the correlation coefficient between the two variables is obtained, and the key operating parameters affecting the performance indicators of the CO2 capture device are determined.

[0070] Step S25, construct a training sample set using the key operating parameter data corresponding to the key operating parameter items affecting the performance indicators of the CO2 capture device and the performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes the key operating parameter data and its corresponding performance indicator data.

[0071] Step S26: training the gated recurrent neural network with samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

[0072] The gated recurrent neural network in this step is used to build a deep learning model between operating parameters and performance indicators. The GRU network is performed from the input layer. After forward information transmission and reverse loss optimization in the GRU layer, the hidden layer weights are continuously updated to learn the rules of the input data. In the process of iterative training, different neurons will provide different contributions to the prediction results. In order to improve the collaborative ability of neurons in the neural network and prevent the occurrence of overfitting, the regularization method Dropout is added after the GRU layer to optimize the model. After Dropout, the data will converge to the fully connected layer and be converted into a one-dimensional vector for output to achieve the prediction of the performance indicators of the CO2 capture device.

[0073] In a detailed embodiment, referring to Figure 3 As shown, the training method of the above machine learning model may include the following steps:

[0074] Step S31, collecting measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis test data.

[0075] The thermometer measurement data include: flue gas inlet and outlet scrubber temperature, flue gas inlet and outlet absorption tower temperature, washing water inlet and outlet scrubber temperature, absorption tower top washing liquid inlet and outlet tower temperature, amine liquid inlet and outlet absorption tower temperature, amine liquid inlet and outlet regeneration tower temperature, flash gas compressor outlet temperature, CO2 gas outlet regeneration tower temperature; pressure measurement data (pressure / differential pressure meter measurement data) include: flue gas inlet and outlet scrubber pressure, flue gas inlet and outlet absorption tower pressure, regeneration tower top pressure, lean liquid flash tank pressure, flash gas compressor outlet pressure, washing tower each section of packing pressure difference, absorption tower each section of packing pressure difference, regeneration tower packing and tower plate pressure difference; flow measurement data include: flue gas inlet and outlet scrubber flow, flue gas inlet and outlet absorption tower flow, washing water inlet and outlet scrubber flow, washing tower supplementary alkali solution flow, amine liquid inlet absorption tower flow, amine liquid inlet regeneration tower flow, amine liquid outlet absorption tower flow, CO2 gas flow; online analytical instrument measurement data include: CO2 content of flue gas entering the scrubber, CO2 content, SO2 content and O2 content of flue gas entering the absorption tower, pH value of washing water entering and leaving the scrubber, pH value and conductivity of amine liquid entering the absorption tower; offline sampling analysis test data include: full component analysis of CO2 content, SO2 content, N2 content, O2 content, particulate matter content and other components of flue gas entering and leaving the scrubber, full component analysis of CO2 content, SO2 content, N2 content, O2 content, particulate matter content and other components of flue gas leaving the absorption tower, COD value (chemical oxygen demand), SS (suspended solids in water quality), pH value of washing water leaving the scrubber, CO2 content, amine concentration, heat-stable salt content, foaming property of amine liquid entering and leaving the absorption tower, CO2 content, amine concentration, heat-stable salt content, foaming property of amine liquid entering and leaving the regeneration tower, and CO2 product purity.

[0076] The temperature measurement data, pressure measurement data, flow measurement data, and online analytical instrument measurement data in this step can be collected in real time through field instruments and DCS remote transmission instrument data; the offline sampling and analysis test data are accumulated through laboratory analysis data.

[0077] Step S32: establishing a full-process simulation model of the CO2 capture device based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

[0078] In this step, refer to Figure 4 As shown, establishing a full-process simulation model of the CO2 capture device based on measurement and analysis data may include the following steps:

[0079] Step S321, based on the process of amine liquid absorbing CO2 in the CO2 capture device, establish an electrolyte property model of the CO2 capture device; wherein the electrolyte property model includes: a phase equilibrium model, an absorption chemical reaction model and a thermal stable salt reaction model.

[0080] The electrolyte physical property model in this step includes adding components and selecting physical property methods. Specifically, add flue gas components, including CO2, O2, N2, H2O, SO2, NO and NO2; add amine liquid components, including MEA and MDEA; add heat-stable salt components, including sodium sulfate, sodium nitrate, sodium nitrite, sodium formate, sodium acetate, sodium oxalate, lactate; add alkali, including NaOH and NAHCO2; select ELECNRTL as the physical property method.

[0081] Among them, adding components includes establishing Henry components, including adding CO2, O2, N2, SO2, NO and NO2 gas phase components as Henry component sets, and using Henry's law combined with the activity coefficient equation to calculate the phase equilibrium model of the system. The absorption chemical reaction model includes the ion equilibrium reaction and parameters of water, the chemical reaction and reaction parameters of CO2 and water, the chemical reaction and reaction parameters of CO2 and MEA, the chemical reaction and reaction parameters of CO2 and MDEA, the hydrolysis reaction and parameters of MEA, the hydrolysis reaction and parameters of MDEA, and the acid-base neutralization reaction and parameters. The establishment of the heat-stable salt reaction model includes the hydrolysis reaction and reaction parameters of all heat-stable salt components. The above electrolyte physical property models will be used in the entire process simulation process.

[0082] Step S322: Selecting an equipment module for drawing a process simulation flow chart based on the electrolyte physical property model.

[0083] The equipment modules in this step may include a rigorous distillation (RADFRAC) model for simulating a tower, a flash (FLASH) model for simulating a tank, a simple or rigorous model for simulating a heat exchanger (HEATER or HEATX), and a pump model (pump), an induced draft fan, and a compressor model (Comp). Preferably, the rigorous distillation (RADFRAC) model for the tower adopts a rate-based model, and the tower internals are set to have a specific surface area of ​​250m 2 / m 3 Metal plate corrugated structured packing.

[0084] Step S323: Based on the measurement and analysis data, logistics data, equipment parameters and operating conditions, the drawn process simulation model is run until the model converges.

[0085] In this step, input data and run the full process simulation model. Among them, logistics data includes the flow rate, temperature and composition of logistics, and logistics includes flue gas, lean liquid, washing water, and washing liquid; equipment parameters include but are not limited to tower packing type and height, heat exchanger outlet temperature, pump and compressor outlet pressure; operating conditions include operating temperature and operating pressure. Input the above data into the logistics and equipment of the process simulation model, and run the adjustment model until it converges correctly.

[0086] In another optional embodiment, also refer to Figure 4 As shown, step S324 may also be included, based on comparing the collected measurement and analysis data with the data simulated by the full process simulation model, and adjusting the model parameters in the electrolyte physical property model based on the comparison results, so that the error between the measurement and analysis data and the simulated data is within a preset range.

[0087] Verifying the full-process simulation model in this step refers to comparing the device data collected in step S31 with the calculated value of the process simulation model, and adjusting and optimizing the model parameters in step S321 so that the deviation between the calculated value of the process simulation model and the measured value of the device site collected in step S31 is less than 5%.

[0088] Step S33: Based on the process principle and engineering experience of the CO2 capture device, determine the key operating parameters and performance indicators of the CO2 capture device. The specific implementation of this step can refer to the above step S21, which will not be repeated here.

[0089] Step S34, obtaining a preset number of key operating parameter data corresponding to key operating parameter items and performance indicator data corresponding to performance indicator items. The data obtained in this step can be the measurement and analysis data in step S31, or the data simulated based on the process simulation model in step S32. In this embodiment, a total of 14,400 sets of data are obtained.

[0090] Step S35: pre-process the key operating parameter data to remove abnormal data. The specific implementation of this step is referred to the above step S23, which will not be repeated here.

[0091] Step S36: perform correlation analysis on the pre-processed key operating parameter data and performance index data to determine the correlation coefficients between different key operating parameter items; and determine the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients. The specific implementation of this step refers to the above step S24 and will not be repeated here.

[0092] Step S37, construct a training sample set with the key operating parameter data corresponding to the key operating parameter items affecting the performance index items of the CO2 capture device and the performance index data corresponding to the performance index items; wherein each sample in the training sample set includes the key operating parameter data and its corresponding performance index data. The specific implementation of this step refers to the above step S25, which will not be repeated here.

[0093] Step S38, training the gated recurrent neural network with samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data. The specific implementation of this step refers to the above step S26, which will not be repeated here.

[0094] Step S39: Use the root mean square error and / or mean absolute error to verify the gap between the predicted value and the true value output by the trained gated recurrent neural network, so as to adjust the model parameters of the gated recurrent neural network.

[0095] In this step, the root mean square error (RMSE) and mean absolute error (MAE) are used to verify the difference between the predicted value and the true value output by the model to indicate the accuracy of the prediction model. The calculation formulas are as follows:

[0096]

[0097]

[0098] Wherein, y^ represents the predicted performance index value of the CO2 capture device, yi represents the actual performance index value of the CO2 capture device, and N represents the number of predictions.

[0099] The above method provided in the embodiment of the present invention, by collecting and utilizing the on-site instrument measurement and on-site analysis data of the device, combined with the process flow simulation model for the CO2 capture device, establishes a deep learning model of the CO2 capture device, and then uses the established deep learning model to predict the utility usage, system water balance, and amine liquid state of the device, and uses the prediction results of the deep learning model to be used in combination with the DCS control system at the device site to intelligently control the operation of the CO2 capture device in real time. The present invention uses the deep learning model of the CO2 capture device to predict key parameters that can be measured in real time and require offline testing and analysis, thereby controlling the operating state of the entire device to a state close to the optimal state, which not only ensures CO2 capture and product purity, but also reduces utility usage and energy consumption, reduces solvent loss, thereby reducing carbon capture costs, and improves the flexibility and convenience of device operation.

[0100] Based on the same inventive concept, an embodiment of the present invention also provides a training device for a machine learning model, referring to Figure 5 As shown, the device may include: a determination module 51, an acquisition module 52, a preprocessing module 53, an analysis module 54, a construction module 55 and a training module 56, and its working principle is as follows:

[0101] The determination module 51 is used to determine the key operating parameter items and performance index items of the CO2 capture device based on the process principle and engineering experience of the CO2 capture device;

[0102] The acquisition module 52 is used to acquire key operation parameter data corresponding to a preset number of key operation parameter items and performance indicator data corresponding to the performance indicator items;

[0103] The preprocessing module 53 is used to preprocess the key operating parameter data to remove the abnormal data contained therein;

[0104] The analysis module 54 is used to perform correlation analysis on the pre-processed key operating parameter data and the performance index data to determine the correlation coefficients between different key operating parameter items; and determine the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients;

[0105] The construction module 55 is used to construct a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance index items of the CO2 capture device and performance index data corresponding to the performance index items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance index data;

[0106] The training module 56 is used to train the gated recurrent neural network using samples in the training sample set, wherein the key operating parameters are optimized and the hidden layer weights are updated in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

[0107] In an alternative embodiment, referring to Figure 5 As shown, the above device may further include: a collection module 57 and a creation module 58;

[0108] The acquisition module 57 is used to acquire the measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis test data;

[0109] The establishing module 58 is used to establish a full-process simulation model of the CO2 capture device based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

[0110] In another optional embodiment, the establishing module 58 is specifically used for:

[0111] Based on the process of amine liquid absorbing CO2 in the CO2 capture device, an electrolyte property model of the CO2 capture device is established; wherein the electrolyte property model includes: a phase equilibrium model, an absorption chemical reaction model and a thermal stable salt reaction model;

[0112] Selecting an equipment module for drawing a process simulation flow chart based on the electrolyte physical property model;

[0113] Based on the measurement and analysis data, logistics data, equipment parameters and operating conditions, the drawn full-process simulation model is run until the model converges.

[0114] In another optional embodiment, the establishing module 58 is further configured to:

[0115] The collected measurement and analysis data are compared with the data simulated by the full-process simulation model, and the model parameters in the electrolyte physical property model are adjusted based on the comparison results so that the error between the measurement and analysis data and the simulated data is within a preset range.

[0116] In an alternative embodiment, referring to Figure 5 As shown, the above-mentioned device may also include: an adjustment module 59, which is used to use the root mean square error and / or the mean absolute error to verify the gap between the predicted value and the true value output by the trained gated recurrent neural network, so as to adjust the model parameters of the gated recurrent neural network.

[0117] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, which, when executed by a processor, implements the training method of the above-mentioned machine learning model.

[0118] Based on the same inventive concept, a computer device is also provided in an embodiment of the present invention, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the training method of the above-mentioned machine learning model when executing the program.

[0119] The principles of the problems solved by the above-mentioned devices, media, and related equipment in the embodiments of the present invention are similar to those of the above-mentioned methods, so their implementation can refer to the implementation of the above-mentioned methods, and the repeated parts will not be repeated.

[0120] Embodiment 2

[0121] In the second embodiment of the present invention, a method for optimizing the process parameters of a flue gas CO2 capture device is provided. Figure 6 As shown, the method may include:

[0122] Step S61: Based on the real-time measurement data of the CO2 capture device acquired in real time and the pre-trained gated recurrent neural network model, predict the key operating parameter data of the offline chemical analysis in the amine liquid system in the CO2 capture device.

[0123] In this step, an intelligent optimization control system for the CO2 capture device can be designed based on the pre-trained gated recurrent neural network model in the above-mentioned embodiment 1 and the DCS control system of the device. The connection between the prediction algorithm and the optimization algorithm and the system is realized to provide guidance for operators to optimize the operating parameters. The optimized operating parameters include but are not limited to optimizing and controlling the usage of public works, which means that the steam usage, circulating water usage and power consumption are minimized when the specified CO2 capture rate and product indicators are achieved; optimizing the water balance control of the system so that the system is in a state of minimal water replenishment or no water replenishment and drainage; optimizing the loss rate of the amine liquid so that the amine loss rate of the device is better than the performance guarantee value; optimizing the state of the amine liquid, including the control of the foaming tendency and the control of the thermally stable salt content.

[0124] Step S62, using the predicted key operating parameter data and real-time measurement data of the offline analysis as the basis for the automatic control and adjustment of the CO2 capture device and the operator to adjust the operating parameters, so as to optimize the process of the CO2 capture device in real time.

[0125] In the specific implementation, it is realized through an intelligent optimization system, which uses but is not limited to tools such as Python programming language, C# programming language, Visual Studio design platform, etc., and can include the following steps:

[0126] Step 1. Convert exe file: Convert the .py script of the prediction model and optimization algorithm program written in Python programming language into .exe file in the interpreter.

[0127] Step 2. Design the operation interface: Use the form in the Visual Studio design platform to design an operation interface with various modules and the corresponding program code, and write and call the .exe file corresponding to the relevant configuration environment.

[0128] Step 3. Read and store prediction results: Connect the Visual Studio design platform to the SQL Serve2015 database to realize the functions of reading and storing prediction result data.

[0129] The designed intelligent optimization control system for the CO2 capture device includes but is not limited to an authorized user login interface, a data management interface, an intelligent algorithm interface, an optimization management interface, and a user management interface.

[0130] (1) User login interface. It is used to verify the information of the login personnel. The operator needs to enter the account and password before logging in. After entering the system, the operator can view and modify some setting parameters in the system.

[0131] (2) Data management interface. It is used to view and process historical and real-time operation data, including online operation data of selected variables and performance indicators and offline analysis data entered by operators. Selected variables include but are not limited to CO2 capture rate, amine solution concentration, circulating amine solution flow rate, CO2 content in circulating amine solution, amine solution inlet temperature of absorption tower, washing liquid flow rate, amine content in washing liquid, and thermal stable salt concentration. Performance indicators include but are not limited to steam consumption, circulating water volume, power consumption, solvent loss rate, system drainage volume, system water replenishment volume, and amine solution defoaming time. The operation change trend of the data is presented in the form of charts.

[0132] (3) Intelligent algorithm management interface. In this interface, you can select the algorithm used to build the prediction model, call the .exe program of the relevant algorithm, and display the calculation results of the corresponding model. It mainly includes two parts. One part is the selection of modeling algorithm. Different prediction algorithms can be selected for display, and the corresponding algorithm adjustable parameters can be changed to meet the needs of changing working conditions on site. In addition to displaying different algorithms, you can also choose whether to perform other related processing on the input variables, such as whether to perform time delay analysis on the data, and whether to perform convolution feature extraction on the decision variables.

[0133] (4) Optimize the management interface. Call NSGA-II to execute the algorithm and display the optimal solution of the Pareto frontier. The output result is the optimal solution with the performance indicators of the CO2 capture device as multiple objectives and the parameters of the corresponding decision variables. The optimized variable parameters are used as the basis for the automatic control and adjustment of the CO2 capture device and the adjustment of the operating parameters by the operator to achieve real-time optimization control.

[0134] The method for optimizing the process parameters of the flue gas CO2 capture device provided in the embodiment of the present invention is to establish a deep learning model of the CO2 capture device by collecting and utilizing the on-site instrument measurement and on-site analysis data of the device, combined with the process flow simulation model for the CO2 capture device, and then use the established deep learning model to predict the utility usage, system water balance, and amine liquid state of the device, and use the prediction results of the deep learning model in combination with the DCS control system at the device site to intelligently control the operation of the CO2 capture device in real time. The present invention uses the deep learning model of the CO2 capture device to predict key parameters that can be measured in real time and require offline testing and analysis, thereby controlling the operating state of the entire device to a state close to the optimal state, which not only ensures CO2 capture and product purity, but also reduces utility usage and energy consumption, reduces solvent loss, thereby reducing carbon capture costs, and improves the flexibility and convenience of device operation.

[0135] Based on the same inventive concept, this method embodiment also provides a device for optimizing the process parameters of the flue gas CO2 capture device, referring to Figure 7 As shown, it may include: a prediction module 71 and an optimization module 72, and its working principle is as follows:

[0136] The prediction module 71 is used to predict the key operating parameter data of the offline chemical analysis in the amine liquid system in the CO2 capture device based on the real-time measurement data of the CO2 capture device acquired in real time and the pre-trained gated recurrent neural network model;

[0137] The optimization module 72 is used to automatically control and adjust the CO2 capture device based on the predicted key operating parameter data of offline chemical analysis and real-time measurement data, and to provide a basis for operators to adjust operating parameters, so as to optimize the process of the CO2 capture device in real time.

[0138] Based on the same inventive concept, a computer-readable storage medium is also provided in the embodiment of the present invention, on which a computer program is stored, and when the program is executed by a processor, the method for optimizing the process parameters of the flue gas CO2 capture device is implemented.

[0139] Based on the same inventive concept, a computer device is also provided in an embodiment of the present invention, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for optimizing the process parameters of the flue gas CO2 capture device is implemented.

[0140] The principles of the problems solved by the above-mentioned devices, media, and related equipment in the embodiments of the present invention are similar to those of the above-mentioned methods, so their implementation can refer to the implementation of the above-mentioned methods, and the repeated parts will not be repeated.

[0141] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0142] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0144] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0145] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for optimizing process parameters of a flue gas CO2 capture device, characterized in that: include: Based on the real-time measurement data of the CO2 capture device acquired in real time and the pre-trained gated recurrent neural network model, predict the key operating parameter data of the offline chemical analysis in the amine liquid system in the CO2 capture device; The predicted key operating parameter data of the offline test analysis and the real-time measurement data are used as the basis for the automatic control and adjustment of the CO2 capture device and the adjustment of the operating parameters by the operator, so as to optimize the process of the CO2 capture device in real time; The gated recurrent neural network model is pre-trained according to the following steps: Based on the process principle and engineering experience of the CO2 capture device, determine the key operating parameters and performance indicators of the CO2 capture device; Acquire a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items; Preprocessing the key operating parameter data to remove abnormal data contained therein; Performing a correlation analysis on the pre-processed key operating parameter data and the performance index data to determine the correlation coefficients between different key operating parameter items; and determining the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients; Constructing a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data; The gated recurrent neural network is trained using samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

2. The method according to claim 1, characterized in that Before acquiring a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items, the method further includes: Collecting measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis and testing data; A full-process simulation model of the CO2 capture device is established based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

3. The method according to claim 2, characterized in that The whole process simulation model of the CO2 capture device is established based on the measurement and analysis data, including: Based on the process of amine liquid absorbing CO2 in the CO2 capture device, an electrolyte property model of the CO2 capture device is established; wherein the electrolyte property model includes: a phase equilibrium model, an absorption chemical reaction model and a thermal stable salt reaction model; Selecting an equipment module for drawing a process simulation flow chart based on the electrolyte physical property model; Based on the measurement and analysis data, logistics data, equipment parameters and operating conditions, the drawn full-process simulation model is run until the model converges.

4. The method according to claim 3, characterized in that Also includes: The collected measurement and analysis data are compared with the data simulated by the full-process simulation model, and the model parameters in the electrolyte physical property model are adjusted based on the comparison results so that the error between the measurement and analysis data and the simulated data is within a preset range.

5. The method according to claim 1, characterized in that Also includes: The root mean square error and / or mean absolute error is used to verify the gap between the predicted value and the true value output by the trained gated recurrent neural network, so as to adjust the model parameters of the gated recurrent neural network.

6. A method for training a machine learning model, characterized in that: include: Based on the process principle and engineering experience of the CO2 capture device, determine the key operating parameters and performance indicators of the CO2 capture device; Acquire a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items; Preprocessing the key operating parameter data to remove abnormal data contained therein; Performing a correlation analysis on the pre-processed key operating parameter data and the performance index data to determine the correlation coefficients between different key operating parameter items; and determining the key operating parameter items that affect the performance index of the CO2 capture device based on the correlation coefficients; Constructing a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data; The gated recurrent neural network is trained using samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

7. The method according to claim 6, characterized in that Before obtaining a preset number of key operating parameter data corresponding to the key operating parameter items and performance indicator data corresponding to the performance indicator items, the method further includes: Collecting measurement and analysis data of the CO2 capture device; wherein the measurement and analysis data includes at least one of the following: temperature measurement data, pressure measurement data, flow measurement data, online analysis instrument measurement data and offline sampling analysis and testing data; A full-process simulation model of the CO2 capture device is established based on the measurement and analysis data, so as to obtain a preset number of key operating parameter data and performance indicator data based on the full-process simulation model.

8. A device for optimizing process parameters of a flue gas CO2 capture device, characterized in that: include: A prediction module, for predicting key operating parameter data of offline chemical analysis in the amine liquid system in the CO2 capture device based on real-time measurement data of the CO2 capture device acquired in real time and a pre-trained gated recurrent neural network model; An optimization module, for automatically controlling and adjusting the CO2 capture device and providing a basis for operators to adjust the operating parameters based on the predicted key operating parameter data obtained by offline chemical analysis and the real-time measurement data, so as to optimize and control the process of the CO2 capture device in real time; A model pre-training module is used to pre-train the gated recurrent neural network model, specifically used to: determine the key operating parameter items and performance index items of the CO2 capture device based on the process principle and engineering experience of the CO2 capture device; obtain a preset number of key operating parameter data corresponding to the key operating parameter items and performance index data corresponding to the performance index items; pre-process the key operating parameter data to eliminate the abnormal data contained therein; A correlation analysis is performed on the preprocessed key operating parameter data and the performance indicator data to determine the correlation coefficients between different key operating parameter items; and based on the correlation coefficients, the key operating parameter items that affect the performance indicators of the CO2 capture device are determined; a training sample set is constructed using the key operating parameter data corresponding to the key operating parameter items that affect the performance indicator items of the CO2 capture device and the performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data; a gated recurrent neural network is trained using the samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss, and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

9. A training device for a machine learning model, characterized in that: include: A determination module, used to determine key operating parameter items and performance index items of the CO2 capture device based on the process principle and engineering experience of the CO2 capture device; An acquisition module, used to acquire a preset number of key operation parameter data corresponding to the key operation parameter items and performance indicator data corresponding to the performance indicator items; A preprocessing module, used for preprocessing the key operating parameter data to remove abnormal data contained therein; An analysis module, used to perform a correlation analysis on the pre-processed key operating parameter data and the performance indicator data to determine the correlation coefficients between different key operating parameter items; and determine the key operating parameter items that affect the performance indicators of the CO2 capture device based on the correlation coefficients; A construction module, used to construct a training sample set with key operating parameter data corresponding to key operating parameter items affecting the performance indicator items of the CO2 capture device and performance indicator data corresponding to the performance indicator items; wherein each sample in the training sample set includes key operating parameter data and its corresponding performance indicator data; A training module is used to train the gated recurrent neural network with samples in the training sample set, wherein the key operating parameters are optimized in the GRU network layer of the gated recurrent neural network through forward information transfer and reverse loss and the hidden layer weights are updated to estimate the model parameters of the performance indicator data corresponding to the key operating parameter data.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the method for optimizing the process parameters of the flue gas CO2 capture device as described in any one of claims 1 to 5, or implements the training method of the machine learning model as described in claim 6 or 7.

11. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements the method for optimizing the process parameters of the flue gas CO2 capture device as described in any one of claims 1 to 5, or implements the training method of the machine learning model as described in claim 6 or 7.

Images