CN111079350A - Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device - Google Patents
Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device Download PDFInfo
- Publication number
- CN111079350A CN111079350A CN201911410141.XA CN201911410141A CN111079350A CN 111079350 A CN111079350 A CN 111079350A CN 201911410141 A CN201911410141 A CN 201911410141A CN 111079350 A CN111079350 A CN 111079350A
- Authority
- CN
- China
- Prior art keywords
- parameters
- data
- tower
- artificial neuron
- key
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a modeling method and a device for the operating performance of a single-tower low-pressure acidic water stripping device, wherein the method comprises the following steps: step S1, collecting full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device, wherein the data source comprises field instrument metering and laboratory analysis data, and data obtained by process simulation of the device; step S2, selecting key parameters, and selecting relevant and actually measurable related parameters aiming at each key parameter to construct an artificial neuron network model between each other; step S3, establishing an artificial neuron network model structure according to the key parameters and the associated parameters; step S4, carrying out normalization processing on the key parameters and the associated parameters participating in modeling; and step S5, performing regression calculation on all parameters in the artificial neuron network model by using an artificial neuron network model training algorithm, so that the prediction of the key parameters by the model is as close as possible to the original data result.
Description
Technical Field
The invention relates to the technical field of chemical engineering, in particular to a modeling method and a modeling device for the operating performance of a single-tower low-pressure acidic water stripping device.
Background
The acidic water stripping device is also called a sewage stripping device, is a common device in the process industries of oil refining, petrochemical industry, oil gas processing and the like, and is used for stripping pollutants such as sulfur, ammonia nitrogen and the like dissolved in water, thereby realizing water quality improvement and avoiding water pollution.
The single-tower low-pressure process is a process commonly used for acidic water stripping at present. As shown in fig. 1, a process diagram of a single-tower low-pressure process in the prior art is shown, hydrogen sulfide, ammonia gas and the like in acidic water/sewage are removed from the tower top through tower bottom heating and tower top condensation, and in the process, a large amount of reboiling heat sources are consumed at the tower bottom.
In a single-tower low-pressure process, acidic water is generated by a plurality of devices at the upstream and is gathered to an acidic water stripping device for treatment, and the characteristic causes the feeding amount and the feeding composition of the acidic water stripping device to fluctuate constantly and the fluctuation range can be large. At the current measurement level of the common industry, the measurement condition of the sour water stripping device is often limited, and parameters such as feeding and operation of the device cannot be completely measured.
In the prior art, parameters such as feeding and operation of the acidic water stripping device are usually measured by adopting a detailed rectifying tower model, however, data requirements of the detailed rectifying tower model include physical properties of fluid, basic information of a tower plate, feeding conditions, reboiling condensation measurement, analysis data of tower top and tower bottom discharge and the like, and the parameter measurement which can be provided in real time in the actual production of the acidic water stripping tower is quite limited, so that the performance description of the acidic water stripping device is difficult to achieve in real time by utilizing the conventional rectifying tower model.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a modeling method and a modeling device for the operating performance of a single-tower low-pressure acidic water stripping device.
To achieve the above and other objects, the present invention provides a method for modeling the operating performance of a single-column low-pressure sour water stripping apparatus, comprising the steps of:
step S1, collecting full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device, wherein the data source comprises field instrument measurement and laboratory analysis data, and the data obtained by process simulation of the device is also used for supplementing data which is difficult to measure in large quantities on the field;
step S2, selecting key parameters of the single-tower acidic water stripping device, and selecting relevant parameters which are related to the key parameters and can be actually measured according to the key parameters, wherein the relevant parameters are used for constructing artificial neuron network mathematical models;
step S3, establishing the structure of the artificial neuron network model according to the selected key parameters and the associated parameters;
step S4, carrying out normalization processing on the key parameters and the associated parameters participating in modeling;
and step S5, performing regression calculation on all parameters in the established artificial neuron network model by using an artificial neuron network model training algorithm, so that the prediction of the key parameters by the artificial neuron network model is as close as possible to the original data result.
Preferably, in step S1, the tower wide operation and analysis data includes:
environmental data including, but not limited to, atmospheric temperature and pressure;
basic column parameters including, but not limited to, number of trays, packing conditions, reboiling pattern, feed location;
distillation target parameters including but not limited to the key component requirements of the bottom discharge purified water and the top discharge acid gas;
operating parameters of the column including, but not limited to, flow rate, temperature, composition analysis of the feed sour water; temperature and pressure distribution in the tower; overhead reflux flow and temperature; overhead acid gas flow; analyzing the flow and composition of purified water at the bottom of the tower; the dosage of a reboiling heat source at the bottom of the tower;
column operating limitations parameters including, but not limited to, feed load variation range; overhead condenser and bottoms reboiler duty ranges.
Preferably, in step S1, data obtained by field instrumentation and laboratory analysis are subjected to corresponding data correction processing based on the principle of material and energy balance.
Preferably, in step S1, a process flow simulation model is built for the single-tower low-pressure sour water stripping apparatus, and the process flow simulation model is used to adjust model parameters within a feasible range of key operating parameters to repeatedly perform simulation calculation, so as to obtain data that is difficult to measure on site in large quantities, and for the process flow simulation model building, data comparison is performed with multi-condition data obtained by field measurement and laboratory analysis, and a system deviation between the process flow simulation data and the field measurement data is determined, so as to ensure that a calculation result of the process flow model is close to an actual measurement result.
Preferably, the key parameters include, but are not limited to, inlet sour water key component content, purified water key component content, overhead sour gas key component content, reboil heat duty, and condenser duty.
Preferably, in step S3, the artificial neuron network model adopts an artificial neuron network of a forward-propagation structure.
Preferably, in step S3, the model structure parameters to be established include, but are not limited to, the neuron number of the input layer, the hidden layer number and the neuron number.
Preferably, in step S4, all the data collected on-site and supplemented by the process simulation model are normalized.
Preferably, in step S5, a back propagation training algorithm is used to find the minimum square difference between the predicted value and the actual measured value of the artificial neuron network by adjusting the weight coefficient W and the bias coefficient b and using a gradient descent optimization method.
In order to achieve the above object, the present invention further provides a modeling apparatus for the operating performance of a single-tower low-pressure sour water stripping apparatus, comprising:
the data acquisition unit is used for acquiring full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device, and the data source of the data acquisition unit comprises field instrument metering and laboratory analysis data and data obtained by carrying out process simulation on the device so as to supplement data which are difficult to measure in large quantities on the field;
the parameter selection unit is used for selecting key parameters of the single-tower acidic water stripping device, selecting relevant parameters which are related to the key parameters and can be actually measured according to the key parameters, and constructing an artificial neuron network mathematical model between the key parameters and the actually measured relevant parameters;
the model building unit is used for building the structure of the artificial neuron network model according to the selected key parameters and the associated parameters;
the normalization processing unit is used for performing normalization processing on the key parameters and the associated parameters participating in modeling;
and the model training unit is used for performing regression calculation on all parameters in the established artificial neuron network model by utilizing an artificial neuron network model training algorithm so that the prediction of the key parameters by the artificial neuron network model is as close to the original data result as possible.
Compared with the prior art, the modeling method and the device for the operating performance of the single-tower low-pressure acidic water stripping device select key parameters of the single-tower low-pressure acidic water stripping device and relevant and practically quantifiable relevant parameters aiming at the key parameters by collecting full-tower operation and analysis data covering the operating fluctuation range of the single-tower low-pressure acidic water stripping device, then establish an artificial neuron network model according to the selected key parameters and relevant parameters, finally utilize an artificial neuron network model training algorithm to carry out regression calculation on all parameters in the established artificial neuron network model so that the prediction of the key parameters by the artificial neuron network model is as close to the original data result as possible, and realize the high robustness performance characterization of the single-tower low-pressure acidic water device by applying the artificial neuron network model to a single-tower low-pressure acidic water process, and a mathematical model basis is provided for optimizing the product quality and energy consumption of single-tower acidic water stripping in real time.
Drawings
FIG. 1 is a process diagram of a single column low pressure process of the prior art;
FIG. 2 is a flow chart of the steps of a modeling method of the operating performance of a single tower low pressure sour water stripping apparatus of the present invention;
fig. 3 is a system architecture diagram of a modeling unit for the operational performance of a single tower low pressure sour water stripping unit of the present invention.
Detailed Description
Other advantages and capabilities of the present invention will be readily apparent to those skilled in the art from the present disclosure by describing the embodiments of the present invention with specific embodiments thereof in conjunction with the accompanying drawings. The invention is capable of other and different embodiments and its several details are capable of modification in various other respects, all without departing from the spirit and scope of the present invention.
FIG. 2 is a flow chart of the steps of a method of modeling the operating performance of a single column low pressure sour water stripping apparatus of the present invention. As shown in FIG. 2, the modeling method for the operation performance of the single-tower low-pressure sour water stripping device comprises the following steps:
and step S1, collecting full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device.
The full-tower operational and analytical data specifically comprises:
1. environmental data: atmospheric temperature and pressure;
2. basic parameters of the column: number of trays, packing conditions, reboiling pattern, feed location;
3. the rectification target parameters are as follows: the key component requirements of purified water discharged from the bottom of the tower and acid gas discharged from the top of the tower are met;
4. operating parameters of the column: analyzing the flow, temperature and composition of the fed acidic water; temperature and pressure distribution in the tower; overhead reflux flow and temperature; overhead acid gas flow; analyzing the flow and composition of purified water at the bottom of the tower; the dosage of a reboiling heat source at the bottom of the tower;
5. operating limiting parameters of the column: feed load variation range; overhead condenser and bottoms reboiler duty ranges.
In a specific embodiment of the present invention, the data acquisition may be implemented by the following two methods: one method is obtained by accumulation of actual production metering instruments and analysis data; another method is to acquire a large amount of data by building a flow simulation model of the device and adjusting model parameters within a feasible range. It should be noted here that, data obtained by analyzing actual production metering instruments and laboratories need to be corrected and processed correspondingly based on the principle of material and energy balance; data obtained by establishing a process simulation model requires field metrology and analytical data to be supplemented to correct the model data to ensure that the model data is sufficiently close to the actual metrology results.
It should be noted here that in application, it is generally necessary to run the process simulation model according to the actual data (to obtain enough data points for training the ANN model because the actual data is difficult to collect enough), and then correct the accuracy of the process simulation model with the actual data, but if the actual data is enough, the correction may not be performed.
Step S2, selecting key parameters (namely target parameters of the artificial neuron network model) of the single-tower sour water stripping device, and selecting parameters (namely related parameters) which are related to the key parameters and can be actually measured for constructing the artificial neuron network mathematical model among the key parameters.
In the specific embodiment of the present invention, the key parameters include, but are not limited to, the content of key components such as ammonia nitrogen or hydrogen sulfide in the feed acidic water, the content of key components in the purified water (i.e., the bottom discharge purified water in step S1), the content of key components in the overhead acidic gas, the reboiling heat load and the condenser load (i.e., the overhead condenser and the bottom reboiler in step S1), and the like. The key parameters play an important role in evaluating the purification effect and the energy consumption level of the device, and analysis data and the like are difficult to measure in real time, so that the method utilizes the artificial neuron network model to predict the key parameters, and can help the device to control the production performance
It is to be noted that the metering levels and the operating fluctuation conditions of different devices are different, so that the key parameters need to be selected according to actual needs
Each key variable can be associated by selecting various different related parameter combinations, and taking the key parameter of the ammonia nitrogen content of the fed acidic water as an example, the associated feed flow (F) can be consideredInlet) And temperature (T)Inlet) Column top temperature (T)Ovhd) Critical tray temperature (T)KeyPlate) Reflux flow (F)Rflx) And temperature (T)Reflx) And amount of reboiled steam (Q)Steam) Gas phase output at the top of the column (V)Top) And the like.
Wherein the content of the first and second substances,
representing the ammonia nitrogen content of the inlet acid water, and NN representing an artificial neuron network model.
In this embodiment, the correlation variables can be increased, such as increasing the atmospheric pressure and the bottom temperature, or decreasing the temperature of the key plate, etc., and the ammonia nitrogen level of the inlet acid water can be evaluated, but the accuracy is different.
In general, the higher the relevance and the larger the number of the parameters participating in the relevance to the key parameters, the better the prediction accuracy of the key parameters.
And step S3, establishing the structure of the artificial neuron network model according to the selected key parameters and the associated parameters.
The artificial neuron network model is also called a neural network or a neural network, and for a single-tower acidic water stripping process, the artificial neuron network with a forward-transfer structure (Feed-forward networks) is adopted in the invention, and the mathematical expression of the model is as follows:
aj＝f1(Wi,jxi+bj)
yk＝f2(Wj,kaj+bk)
wherein x isiIs the input of the neuron, i.e., the associated parameter selected in step S2, ajAre intermediate layer neurons, ykFor outputting the layer result, i.e., the result of the calculation of the key variable selected in step S2 by the neural network model, W and b are the weight coefficient and bias coefficient corresponding to each neuron of each layer, respectively, f1And f2For the propagation function, in the present invention, the propagation function selects a Sigmoid type function, namely:
z represents a propagation function f1And f2A numerical value is input.
In the specific embodiment of the invention, the model structure parameters to be established include the number of neurons in the input layer, the number of hidden layers, the number of neurons and the like, and for the same set of target parameters and associated parameters, different neuron network result models can result in different model output accuracies.
And step S4, carrying out normalization processing on the key parameters and the associated parameters participating in modeling.
In the embodiment of the invention, all the data acquired on site need to be normalized based on the principle of material and energy balance.
Step S5, using an artificial neuron network model training algorithm to perform regression calculation on all parameters (weight coefficient W and bias coefficient b) in the established artificial neuron network model, so that the prediction of the key parameters by the artificial neuron network model is as close to the original data result as possible (namely the data collected in step S1).
In the specific embodiment of the invention, a back propagation training algorithm is adopted, and the algorithm searches the minimum square difference between the predicted value and the actual measured value of the artificial neuron network by adjusting the weight coefficient W and the bias coefficient b and utilizing a gradient descent optimization method:
where N is the number of samples participating in the training, tkThe method is characterized in that a predicted value of an artificial neuron network model is adopted, and F is the minimum evaluation variance of the predicted value and an actually measured value.
The trained model of the invention can be used for predicting each key parameter under different operation condition combinations, and can be used for comprehensively describing the operation and performance conditions of the single-tower sour water stripping device.
FIG. 3 is a system block diagram of a modeling unit for the operational performance of a single-tower low-pressure sour water stripping unit of the present invention. As shown in fig. 3, the modeling apparatus for the operation performance of the single-tower low-pressure sour water stripping apparatus according to the present invention comprises:
and the data acquisition unit 301 is used for acquiring full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device.
The full-tower operational and analytical data specifically comprises:
1. environmental data: atmospheric temperature and pressure;
2. basic parameters of the column: number of trays, packing conditions, reboiling pattern, feed location;
3. the rectification target parameters are as follows: the key component requirements of purified water discharged from the bottom of the tower and acid gas discharged from the top of the tower are met;
4. operating parameters of the column: analyzing the flow, temperature and composition of the fed acidic water; temperature and pressure distribution in the tower; overhead reflux flow and temperature; overhead acid gas flow; analyzing the flow and composition of purified water at the bottom of the tower; the dosage of a reboiling heat source at the bottom of the tower;
5. operating limiting parameters of the column: feed load variation range; overhead condenser and bottoms reboiler duty ranges.
In the embodiment of the present invention, the data acquisition unit 301 may acquire data by the following two methods: one method is obtained by accumulation of actual production metering instruments and analysis data; another method is to acquire a large amount of data by building a flow simulation model of the device and adjusting model parameters within a feasible range. It should be noted here that, data obtained by analyzing actual production metering instruments and laboratories need to be corrected and processed correspondingly based on the principle of material and energy balance; data obtained by establishing a process simulation model requires field metrology and analytical data to be supplemented to correct the model data to ensure that the model data is sufficiently close to the actual metrology results.
The parameter selection unit 302 is configured to select key parameters of the single-tower acid water stripping device (i.e., target parameters of the artificial neuron network model), and select, for each key parameter, a parameter (i.e., associated parameter) related to the key parameter and actually measurable, and construct an artificial neuron network mathematical model between the key parameters and the actual parameters.
In the specific embodiment of the present invention, the key parameters include, but are not limited to, the content of key components such as ammonia nitrogen or hydrogen sulfide in the inlet acidic water, the content of key components in the purified water, the content of key components in the acid gas at the top of the tower, the reboiling heat load, the condenser load, and the like. The key parameters play an important role in evaluating the purification effect and the energy consumption level of the device, and analysis data and the like are difficult to measure in real time, so that the method utilizes the artificial neuron network model to predict the key parameters, and can help the device to control the production performance
It should be noted here that the metering levels and the operating fluctuation conditions of different devices are different, and therefore, the key parameters need to be selected according to actual needs.
Each key variable can be associated by selecting various different related parameter combinations, and taking the key parameter of the ammonia nitrogen content of the fed acidic water as an example, the associated feed flow (F) can be consideredInlet) And temperature (T)Inlet) Column top temperature (T)Ovhd) Critical tray temperature (T)KeyPlate) Reflux flow (F)Rflx) And temperature (T)Reflx) And amount of reboiled steam (Q)Steam) Gas phase output at the top of the column (V)Top) And the like.
Wherein the content of the first and second substances,
representing the ammonia nitrogen content of the inlet acid water, and NN representing an artificial neuron network model.
In this embodiment, the correlation variables can be increased, such as increasing the atmospheric pressure and the bottom temperature, or decreasing the temperature of the key plate, etc., and the ammonia nitrogen level of the inlet acid water can be evaluated, but the accuracy is different.
In general, the higher the relevance and the larger the number of the parameters participating in the relevance to the key parameters, the better the prediction accuracy of the key parameters.
And the model building unit 303 is configured to build a structure of the artificial neuron network model according to the selected key parameter and the associated parameter.
The artificial neuron network model is also called a neural network or a neural network, and for a single-tower acidic water stripping process, the artificial neuron network with a forward-transfer structure (Feed-forward networks) is adopted in the invention, and the mathematical expression of the model is as follows:
aj＝f1(Wi,jxi+bj)
yk＝f2(Wj,kaj+bk)
wherein x isiIs the input of the neuron, i.e., the associated parameter selected in step S2, ajAre intermediate layer neurons, ykFor outputting the layer result, i.e., the result of the calculation of the key variable selected in step S2 by the neural network model, W and b are the weight coefficient and bias coefficient corresponding to each neuron of each layer, respectively, f1And f2For the propagation function, in the present invention, the propagation function selects a Sigmoid type function, namely:
z represents a propagation function f1And f2A numerical value is input.
In the embodiment of the present invention, the model structure parameters that need to be established by the model establishing unit 303 include the number of neurons in the input layer, the number of hidden layers, the number of neurons, and the like, and for the same set of target parameters and associated parameters, different neuron network result models result in different model output accuracies.
And the normalization processing unit 304 is configured to perform normalization processing on the key parameters and the associated parameters participating in modeling.
In an embodiment of the present invention, the normalization processing unit 304 normalizes all the data collected in the field based on the principle of material balance and energy balance.
The model training unit 305 is configured to perform regression calculation on all parameters (weight coefficient W and bias coefficient b) in the established artificial neuron network model by using an artificial neuron network model training algorithm, so that the prediction of the artificial neuron network model on the key parameters is as close as possible to the original data result.
In the embodiment of the present invention, the model training unit 305 adopts a back propagation training algorithm, which uses a gradient descent optimization method to find the least square difference between the predicted value and the actual measured value of the artificial neuron network by adjusting the weight coefficient W and the bias coefficient b:
where N is the number of samples participating in the training, tkThe method is characterized in that a predicted value of an artificial neuron network model is adopted, and F is the minimum evaluation variance of the predicted value and an actually measured value.
Examples
In this embodiment, a single-tower acidic water stripping apparatus containing sulfur and ammonia nitrogen in an enterprise is taken as an example, the content of imported ammonia nitrogen is taken as a key variable to be predicted by a model, a modeling method and an effect are demonstrated, and a modeling process is as follows
1) And establishing a process simulation model of the single-tower acidic water stripping device by using process simulation software (such as AspenHYSYS), and adjusting the model according to the DCS and the analysis data which are actually measured so that the result of the process simulation model is close to the actual result.
2) Selecting key parameters and associated parameters:
key parameters are as follows: ammonia nitrogen content of inlet acidic water
And (3) correlation parameters: temperature T of the feedInletTemperature T at the top of the columnOvhdCritical tray temperature TKeyPlateReflux temperature TReflxFeed rate FInletAmount of reflux FRflxGas phase output V from the tower topTopAnd the amount of reboiled steam QSteam。
3) Defining adjustment or fluctuation range for operation variables according to actual production limiting conditions, such as setting boundaries for feeding temperature and ammonia nitrogen content of acidic water according to actual historical fluctuation, setting boundaries for feeding flow according to design fluctuation allowable range, setting boundaries for ammonia nitrogen content of purified water according to environmental protection requirements and historical analysis fluctuation range, and the like.
4) And adjusting parameters of the process simulation model in the variable fluctuation range, and repeatedly executing simulation calculation to obtain about 2800 groups of different operation data of the device.
5) Establishing the structure of the artificial neuron network model according to the key parameters and the associated parameters selected in the step 1): one intermediate layer containing 6 neurons was selected.
6) And carrying out normalized data processing on the associated parameters and the key parameters.
7) And training the model by using a back propagation algorithm, and adjusting the weight coefficient and the bias coefficient to ensure that the prediction of the model on the key parameters is closest to the original data.
8) In 2800 training samples, the neuron model predicted the ammonia nitrogen content of the feed to the sour water stripper with an average error of 1.53% and a maximum error of 5.49%.
9) The trained model is applied to actual production, and the numerical values of the key parameters can be pre-judged by using the running values of the associated parameters.
In summary, the modeling method and device for the operating performance of the single-tower low-pressure acidic water stripping device of the invention selects the key parameters of the single-tower low-pressure acidic water stripping device and the relevant parameters which are relevant to the key parameters and can be actually measured by collecting the whole-tower operation and analysis data covering the operating fluctuation range of the single-tower low-pressure acidic water stripping device, then establishes the artificial neuron network model according to the selected key parameters and relevant parameters, finally uses the artificial neuron network model training algorithm to regress all the parameters in the established artificial neuron network model, so that the prediction of the key parameters by the artificial neuron network model is as close to the original data result as possible, the invention realizes the high robustness performance characterization of the single-tower low-pressure acidic water device by applying the artificial neuron network model to the single-tower low-pressure acidic water process, and a mathematical model basis is provided for optimizing the product quality and energy consumption of single-tower acidic water stripping in real time.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Modifications and variations can be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the present invention. Therefore, the scope of the invention should be determined from the following claims.
Claims (10)
1. A modeling method for the operating performance of a single-tower low-pressure sour water stripping device comprises the following steps:
step S1, collecting full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device, wherein the data source comprises field instrument measurement and laboratory analysis data, and the data obtained by process simulation of the device is also used for supplementing data which is difficult to measure in large quantities on the field;
step S2, selecting key parameters of the single-tower acidic water stripping device, and selecting relevant parameters which are related to the key parameters and can be actually measured according to the key parameters, wherein the relevant parameters are used for constructing artificial neuron network mathematical models;
step S3, establishing the structure of the artificial neuron network model according to the selected key parameters and the associated parameters;
step S4, carrying out normalization processing on the key parameters and the associated parameters participating in modeling;
and step S5, performing regression calculation on all parameters in the established artificial neuron network model by using an artificial neuron network model training algorithm, so that the prediction of the key parameters by the artificial neuron network model is as close as possible to the original data result.
2. The method of claim 1 wherein the full column operating and analytical data in step S1 includes:
environmental data including, but not limited to, atmospheric temperature and pressure;
basic column parameters including, but not limited to, number of trays, packing conditions, reboiling pattern, feed location;
distillation target parameters including but not limited to the key component requirements of the bottom discharge purified water and the top discharge acid gas;
operating parameters of the column including, but not limited to, flow rate, temperature, composition analysis of the feed sour water; temperature and pressure distribution in the tower; overhead reflux flow and temperature; overhead acid gas flow; analyzing the flow and composition of purified water at the bottom of the tower; the dosage of a reboiling heat source at the bottom of the tower;
column operating limitations parameters including, but not limited to, feed load variation range; overhead condenser and bottoms reboiler duty ranges.
3. A method for modeling the operational performance of a single column low pressure sour water stripping apparatus according to claim 1, wherein the data obtained by field instrumentation and laboratory analysis is subjected to a corresponding data correction process based on the principles of material and energy balance in step S1.
4. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: in step S1, a process flow simulation model is established for the single-tower low-pressure sour water stripping apparatus, and model parameters are adjusted within a feasible range of key operating parameters by using the process flow simulation model to repeatedly perform simulation calculation, so that data which are difficult to measure on site in large quantities are acquired in a large quantity manner; the process flow simulation model is established by comparing data with multi-working condition data obtained by field measurement and laboratory analysis, and system deviation between the process flow simulation data and the field measurement data is determined, so that the calculated result of the process flow model is close to the actual measurement result.
5. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: the key parameters include, but are not limited to, inlet sour water key component content, purified water key component content, overhead sour gas key component content, reboil heat duty, and condenser duty.
6. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: in step S3, the artificial neuron network model employs an artificial neuron network of a forward-propagation structure.
7. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: in step S3, the model structure parameters to be established include, but are not limited to, the number of neurons in the input layer, the number of hidden layers, and the number of neurons.
8. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: in step S4, all the field data that remain after data correction and the supplemental data obtained based on the process simulation model are normalized.
9. A method of modeling the operating performance of a single column low pressure sour water stripping plant as claimed in claim 1 wherein: in step S5, a back propagation training algorithm is used to find the minimum square difference between the predicted value and the actual measured value of the artificial neuron network by adjusting the weight coefficient W and the bias coefficient b and using a gradient descent optimization method.
10. A modeling apparatus for single tower low pressure sour water stripping apparatus operating performance, comprising:
the data acquisition unit is used for acquiring full-tower operation and analysis data covering the operation fluctuation range of the single-tower low-pressure acidic water stripping device, and the data source of the data acquisition unit comprises field instrument metering and laboratory analysis data and data obtained by carrying out process simulation on the device so as to supplement data which are difficult to measure in large quantities on the field;
the parameter selection unit is used for selecting key parameters of the single-tower acidic water stripping device, selecting relevant parameters which are related to the key parameters and can be actually measured according to the key parameters, and constructing an artificial neuron network mathematical model between the key parameters and the actually measured relevant parameters;
the model building unit is used for building the structure of the artificial neuron network model according to the selected key parameters and the associated parameters;
the normalization processing unit is used for performing normalization processing on the key parameters and the associated parameters participating in modeling;
and the model training unit is used for performing regression calculation on all parameters in the established artificial neuron network model by utilizing an artificial neuron network model training algorithm so that the prediction of the key parameters by the artificial neuron network model is as close to the original data result as possible.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911410141.XA CN111079350A (en) | 2019-12-31 | 2019-12-31 | Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911410141.XA CN111079350A (en) | 2019-12-31 | 2019-12-31 | Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111079350A true CN111079350A (en) | 2020-04-28 |
Family
ID=70321048
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911410141.XA Pending CN111079350A (en) | 2019-12-31 | 2019-12-31 | Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111079350A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5385646A (en) * | 1993-09-03 | 1995-01-31 | Farmland Industries, Inc. | Method of treating chemical process effluent |
| US20030234218A1 (en) * | 2002-06-21 | 2003-12-25 | H2L Co., Ltd. | System and method for AI controlling waste-water treatment by neural network and back-propagation algorithm |
| US20050047981A1 (en) * | 2000-08-10 | 2005-03-03 | Hirofumi Kikkawa | Process and apparatus for treating ammonia-containing waste water |
| US20060073013A1 (en) * | 2004-09-10 | 2006-04-06 | Emigholz Kenneth F | Application of abnormal event detection technology to fluidized catalytic cracking unit |
| CN1762525A (en) * | 2005-10-14 | 2006-04-26 | 清华大学 | Rectification tower automatic control and optimization method |
| US20150034553A1 (en) * | 2011-07-26 | 2015-02-05 | General Electric Company | Wastewater treatment plant online monitoring and control |
| CN106477792A (en) * | 2016-09-26 | 2017-03-08 | 中国石油化工股份有限公司 | A kind of method and device for stripping process petrochemical industry sour water nothing side line single column negative pressure |
| CN108204997A (en) * | 2016-12-20 | 2018-06-26 | 中国石油天然气股份有限公司 | Normal line oil flash point online soft sensor method |
| CN110069889A (en) * | 2019-05-06 | 2019-07-30 | 山东华星石油化工集团有限公司 | A kind of diesel hydrotreating unit stripper steam adaptation method based on way of escape measuring and calculating |
-
2019
- 2019-12-31 CN CN201911410141.XA patent/CN111079350A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5385646A (en) * | 1993-09-03 | 1995-01-31 | Farmland Industries, Inc. | Method of treating chemical process effluent |
| US20050047981A1 (en) * | 2000-08-10 | 2005-03-03 | Hirofumi Kikkawa | Process and apparatus for treating ammonia-containing waste water |
| US20030234218A1 (en) * | 2002-06-21 | 2003-12-25 | H2L Co., Ltd. | System and method for AI controlling waste-water treatment by neural network and back-propagation algorithm |
| US20060073013A1 (en) * | 2004-09-10 | 2006-04-06 | Emigholz Kenneth F | Application of abnormal event detection technology to fluidized catalytic cracking unit |
| CN1762525A (en) * | 2005-10-14 | 2006-04-26 | 清华大学 | Rectification tower automatic control and optimization method |
| US20150034553A1 (en) * | 2011-07-26 | 2015-02-05 | General Electric Company | Wastewater treatment plant online monitoring and control |
| CN106477792A (en) * | 2016-09-26 | 2017-03-08 | 中国石油化工股份有限公司 | A kind of method and device for stripping process petrochemical industry sour water nothing side line single column negative pressure |
| CN108204997A (en) * | 2016-12-20 | 2018-06-26 | 中国石油天然气股份有限公司 | Normal line oil flash point online soft sensor method |
| CN110069889A (en) * | 2019-05-06 | 2019-07-30 | 山东华星石油化工集团有限公司 | A kind of diesel hydrotreating unit stripper steam adaptation method based on way of escape measuring and calculating |
Non-Patent Citations (8)
| Title |
|---|
| 刘世达;王海燕;任龙;: "Pro/Ⅱ软件在酸性水单塔汽提装置优化操作中应用" * |
| 刘云波,刘艳升,沈复,高凌霄,张丕烈: "人工神经网络用于精馏塔产品质量预测的研究" * |
| 李章平,孙秋荣: "Apsen Plus软件在含硫废水汽提工艺设计中的应用" * |
| 李菁菁;: "酸性水单塔低压汽提工艺能耗分析及节能措施" * |
| 王正;汪建华;: "青岛炼化酸性水汽提装置流程模拟与优化" * |
| 王洪海;张玉珍;李月;耿海腾;李春利;: "基于正交设计和神经网络的萃取精馏工艺优化" * |
| 连晓峰;刘载文;刘萌;王磊;王小艺;: "三相生物流化床味精废水处理的优化控制研究" * |
| 高明;薄徳臣；陈建兵;: "酸性水汽提装置研析" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10739752B2 (en) | Computer system and method for causality analysis using hybrid first-principles and inferential model | |
| US6088630A (en) | Automatic control system for unit operation | |
| CN111142494B (en) | Intelligent control method and system for amine liquid regeneration device | |
| RU2280270C2 (en) | On-line calibration method | |
| CN107403196B (en) | Method for predicting concentration of butane at bottom of debutanizer by instant learning modeling based on spectral clustering analysis | |
| Li et al. | Time-delay identification in dynamic processes with disturbance via correlation analysis | |
| CN108595892A (en) | Soft-measuring modeling method based on time difference model | |
| CN113420500B (en) | Intelligent atmospheric and vacuum system | |
| Al Jlibawi et al. | Optimization of distribution control system in oil refinery by applying hybrid machine learning techniques | |
| JP5110891B2 (en) | Statistical prediction method and apparatus for influent water quality in water treatment facilities | |
| CN201330211Y (en) | Working parameter self-optimizing simulation system for sewage treatment plant | |
| CN111079350A (en) | Modeling method and device for operating performance of single-tower low-pressure acidic water stripping device | |
| NASSER et al. | Online composition prediction of a debutanizer column using artificial neural network | |
| Bo et al. | The application of neural network soft sensor technology to an advanced control system of distillation operation | |
| CN114970375B (en) | Rectification process monitoring method based on real-time sampling data | |
| CN115631804A (en) | Method for predicting outlet concentration of sodium aluminate solution in evaporation process based on data coordination | |
| Zhang et al. | Soft sensor for cobalt oxalate synthesis process in cobalt hydrometallurgy based on hybrid model | |
| Jiang et al. | Optimal product quality control in a hydrocracking fractionator with process simulation approaches | |
| Udugama et al. | Cost competitive “soft sensor” for determining product recovery in industrial methanol | |
| CN111781821B (en) | Parameter self-tuning method of SISO (SISO) compact-format model-free controller based on Attention mechanism cyclic neural network | |
| Juuso et al. | Intelligent analysers and dynamic simulation in a biological water treatment process | |
| Lüthje et al. | Adaptive learning of hybrid models for nonlinear model predictive control of distillation columns | |
| CN113420499B (en) | Physical logic reconstruction method for atmospheric and vacuum distillation unit | |
| Fieg | Composition control of distillation columns with a sidestream by using gas chromatographs | |
| Ebnali et al. | Artificial intelligence for inferential control of crude oil stripping process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Lou Yuhang Inventor after: Zhang Nan Inventor after: Li Guanghui Inventor after: Rao Pinhua Inventor after: Lu Jiaqi Inventor before: Lou Yuhang Inventor before: Zhang Nan |
|
| CB03 | Change of inventor or designer information |