CN113393905B - Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system - Google Patents
Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system Download PDFInfo
- Publication number
- CN113393905B CN113393905B CN202110617183.1A CN202110617183A CN113393905B CN 113393905 B CN113393905 B CN 113393905B CN 202110617183 A CN202110617183 A CN 202110617183A CN 113393905 B CN113393905 B CN 113393905B
- Authority
- CN
- China
- Prior art keywords
- absorption tower
- state
- outlet
- inlet
- time domain
- 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.)
- Active
Links
- 238000010521 absorption reaction Methods 0.000 title claims abstract description 146
- 238000005259 measurement Methods 0.000 title claims abstract description 33
- 239000000126 substance Substances 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 title claims abstract description 18
- 238000005096 rolling process Methods 0.000 claims abstract description 49
- 238000013528 artificial neural network Methods 0.000 claims abstract description 46
- 239000007788 liquid Substances 0.000 claims description 52
- 239000007789 gas Substances 0.000 claims description 32
- 239000011159 matrix material Substances 0.000 claims description 31
- 238000005070 sampling Methods 0.000 claims description 22
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 claims description 21
- 239000003546 flue gas Substances 0.000 claims description 21
- 238000000691 measurement method Methods 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 5
- 238000000746 purification Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 abstract description 10
- 229910052799 carbon Inorganic materials 0.000 abstract description 10
- 239000006096 absorbing agent Substances 0.000 abstract description 8
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 238000010926 purge Methods 0.000 abstract description 3
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 16
- 238000010586 diagram Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 9
- 229910002092 carbon dioxide Inorganic materials 0.000 description 8
- 239000001569 carbon dioxide Substances 0.000 description 8
- 239000002904 solvent Substances 0.000 description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005262 decarbonization Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C10/00—Computational theoretical chemistry, i.e. ICT specially adapted for theoretical aspects of quantum chemistry, molecular mechanics, molecular dynamics or the like
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/46—Removing components of defined structure
- B01D53/62—Carbon oxides
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/74—General processes for purification of waste gases; Apparatus or devices specially adapted therefor
- B01D53/77—Liquid phase processes
- B01D53/78—Liquid phase processes with gas-liquid contact
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
- Y02A50/20—Air quality improvement or preservation, e.g. vehicle emission control or emission reduction by using catalytic converters
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02C—CAPTURE, STORAGE, SEQUESTRATION OR DISPOSAL OF GREENHOUSE GASES [GHG]
- Y02C20/00—Capture or disposal of greenhouse gases
- Y02C20/40—Capture or disposal of greenhouse gases of CO2
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Analysis (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Computational Mathematics (AREA)
- Oil, Petroleum & Natural Gas (AREA)
- Analytical Chemistry (AREA)
- Mathematical Optimization (AREA)
- Artificial Intelligence (AREA)
- Pure & Applied Mathematics (AREA)
- General Chemical & Material Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Algebra (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Biophysics (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Crystallography & Structural Chemistry (AREA)
- Treating Waste Gases (AREA)
Abstract
The invention discloses a chemical absorption method of CO 2 Dynamic robust soft measurement system and method of trapping system, mainly aiming at CO 2 Operating parameters in the capture system that are difficult to measure accurately in real time, such as absorber inlet lean load and absorber outlet purge gas CO 2 The concentration is measured softly, so that the system is ensured to operate efficiently and stably, and support is provided for implementing dynamic control. The invention adopts BP neural network technology, obtains a high-precision simplified model of the carbon trapping system based on dynamic data identification generated by the operation of a chemical absorption carbon trapping mechanism model, and realizes reliable soft measurement of a target variable when unknown interference, noise, model mismatch and partial parameter measurement faults by using a robust on-line monitoring method based on a rolling time domain on the basis.
Description
Technical Field
The invention relates to the technical field of online soft measurement methods of chemical absorption carbon capture systems, in particular to a chemical absorption CO based on neural network and rolling time domain estimation 2 A dynamic robust soft measurement method for a trapping system.
Background
CO based on chemical absorption 2 Trapping is the most mature and promising technology for decarbonization of energy and industrial processes. However, due to the reduced performance of the solvent and corrosion of the equipment, it is still difficult to maintain a stable and efficient operation of the system over a long period of time. In addition, flexible control of the carbon capture system is also highly desirable to ensure that the system can timely adapt to upstream flue gas changes and CO 2 Product(s)Is not limited to the above-mentioned requirements.
On-line monitoring of chemical absorption CO 2 The key parameters in the trapping process can be used for knowing the running state of the system and laying a foundation for the design of the control system. However, due to measurement technology limitations, some important parameters in the carbon capture system, such as CO 2 Concentration, lean load, etc., particularly under complex operating conditions, are difficult to measure in real-time, accurately, reliably, and economically. Therefore, the invention provides a chemical absorption CO based on neural network and rolling time domain estimation 2 A dynamic robust soft measurement method of a trapping system,
disclosure of Invention
The invention aims to solve the technical problem of providing a chemical absorption CO based on a neural network and rolling time domain estimation 2 Dynamic robust soft measurement system and method of trapping system can realize CO absorption based on solvent under the condition of complex operation such as noise, unknown interference, fault and the like 2 Dynamic real-time robust estimation of key parameters in a trapping system.
In order to solve the technical problems, the invention adopts the following technical scheme:
chemical absorption CO based on neural network and rolling time domain estimation 2 A trapping system dynamic robust soft measurement system, comprising: rolling time domain estimator (1), unscented Kalman filter estimator (2), BP neural network simplified model (3), chemical absorption CO 2 A trapping system model (4); the BP neural network simplified model (3) is chemical absorption CO 2 Capturing a simplified model among core parameters of the overall system model (4) and updating an equation f (x) as a state of the rolling time domain estimator (1); the output quantity of the rolling time domain estimator (1) to the BP neural network simplified model (3) is the assumed value { x ] of the arrival cost function to the state quantity t0 -a }; the rolling time domain estimator (1) receives the output { x } from the BP neural network reduced model (3) t0+1 -for calculating the arrival cost Φ T Obtaining the estimated value of the state quantity at the next moment by minimizing the arrival costThe unscented Kalman filter estimator (2) rolls the estimated value of the state quantity of the time domain estimator (1) according to the corresponding time>And error covariance matrix pi T-N-1 To calculate and update an error covariance matrix pi for the next instant of the rolling horizon estimator (1) T-N 。
Wherein the chemical absorption of CO 2 Capturing the overall system model (4) core parameters includes: lean liquid load at inlet of absorption tower and purified gas CO at outlet of absorption tower at current moment 2 Concentration, purification gas temperature at the outlet of the absorption tower, rich liquid temperature at the outlet of the absorption tower, lean liquid flow at the inlet of the absorption tower and flue gas CO at the inlet of the absorption tower 2 Concentration, absorption tower inlet flue gas flow, reboiler temperature, absorption tower inlet lean solution load at next moment, absorption tower outlet purified gas CO 2 Concentration, temperature of purified gas at the outlet of the absorption tower and temperature of rich liquid at the outlet of the absorption tower.
Chemical absorption CO based on neural network and rolling time domain estimation 2 The dynamic robust soft measurement method of the trapping system is characterized by comprising the following steps of:
step (1), selecting lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower 2 The concentration is the to-be-estimated value, the temperature of purified gas at the outlet of the absorption tower and the temperature of rich liquid at the outlet of the absorption tower are selected as the measurement values, and the flow of lean liquid at the inlet of the absorption tower and the CO of flue gas at the inlet of the absorption tower are selected 2 The concentration, the flue gas flow at the inlet of the absorption tower and the temperature of the reboiler are used as input quantities, and initial values are set;
step (2), setting the time domain window size N, the state covariance Q, the measurement covariance R and the state error covariance matrix pi in the rolling time domain estimator (1) 0 And state quantity constraint values and prior distribution factors beta of state covariance Qu, measurement covariance Ru, scale factors alpha, scale parameters k and x in the unscented Kalman filter estimator (2);
step (3), assuming that the state variable x in the BP neural network simplified model (3) contains Gaussian white noise W (k), and the observed value y contains Gaussian white noise V (k), forming system nonlinearity, such as formula (1):
wherein x is k Represents k moment of lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower 2 Vector of concentration, purification gas temperature at the outlet of the absorption tower and rich liquid temperature at the outlet of the absorption tower; u (u) k Represents k moment of flow of lean liquid at the inlet of the absorption tower and CO at the inlet of the absorption tower 2 Vector of concentration, absorption tower inlet flue gas flow and reboiler temperature; w (w) k And v k The state noise and the measurement noise at the moment k are respectively; f (x, u) is a BP neural network simplified model (3); because the temperature of the purified gas at the outlet of the absorption tower and the temperature of the rich liquid at the outlet of the absorption tower are measurable and serve as state quantity, L is linear transformation with the diagonal angle of 1;
step (4), when T is less than or equal to N, obtaining a state estimation value of the current moment of the rolling time domain estimator (1) by solving an objective function phi for minimizing full information estimation, such as a formula (2), wherein a state error covariance matrix defaults to pi 0 ：
Step (5), when T > N, obtaining a state estimation value of the current moment of the rolling time domain estimator (1) by solving an objective function phi of the minimum approximate estimation as shown in a formula (3):
step (6), using UT conversion by unscented Kalman filter estimator (2), estimating value from state of T-N-1 timeSum error covariance matrix pi T-N-1 In this process, the state estimation value at the time T-N-1 is recorded/>Is->Error covariance matrix pi at time T-N-1 T-N-1 Is P xp A symmetrical sampling method is adopted to generate a sigma point set { χ } i -as in equation (4):
where n is the state dimension, k=2n+1 is the number of sigma sampling points, W i (i=0, 1 … … K-1) is the weight of the sigma point set,and->The weights of the mean and covariance, respectively; λ=α 2 (n+k) -n is used to adjust sigma point to +.>Is a distance of (2);
step (7), calculating a prediction state sampling point of sigma sampling points of the BP neural network simplified model (3)As in formula (5):
step (8), sampling points according to the prediction stateAnd corresponding weight W i Calculating the predicted state mean->Sum covariance P k+1|k As in formula (6):
step (9), sampling points according to the prediction stateCalculating measurement sampling points of the BP neural network simplified model (3), such as a formula (7):
step (10), calculating the mean and covariance of the predicted measurements, as in equation (8):
P y to observe the variance matrix, P xy Is the covariance matrix of the observations and the state vector;
step (11), calculating and updating an error covariance matrix pi of the next moment of the rolling time domain estimator (1) T-N As in formula (9):
and (12) repeatedly executing the steps (5) to (11) in the following period, updating to obtain an error covariance matrix of the new moment, and then solving a state estimation value of the next moment by using the rolling time domain estimator (1).
The invention has the beneficial effects that:
the invention discloses a chemical absorption CO based on neural network and rolling time domain estimation 2 Trapping system dynamic robust soft measurementSystem and method mainly aiming at CO 2 Operating parameters in the capture system that are difficult to measure accurately in real time, such as absorber inlet lean load and absorber outlet purge gas CO 2 The concentration is measured softly, so that the system is ensured to operate efficiently and stably, and support is provided for implementing dynamic control. The invention adopts BP neural network technology, obtains a high-precision simplified model of the carbon trapping system based on dynamic data identification generated by the operation of a chemical absorption carbon trapping mechanism model, uses a robust on-line monitoring method based on a rolling time domain on the basis, and can realize CO absorption based on a solvent under the complex operation conditions of noise, unknown interference, model mismatch, partial parameter measurement faults and the like 2 More accurate robust estimation of key parameters in the trapping system.
Drawings
FIG. 1 is a schematic diagram of the system of the present invention.
FIG. 2 is a schematic illustration of the chemical absorption of CO according to the invention 2 Schematic flow diagram of trapping system.
Fig. 3 (a) is a schematic diagram showing the effect of the present invention on estimating the lean liquid load at the inlet of the absorber when noise interference is added.
FIG. 3 (b) is a schematic diagram showing the effect of the present invention on estimating the carbon dioxide concentration at the outlet of the absorber when noise interference is added.
Fig. 4 (a) is a schematic diagram showing the effect of the present invention on estimating the lean liquid load at the inlet of the absorption tower when noise interference is added and the temperature of the purified air at the outlet of the absorption tower is measured at a fault.
Fig. 4 (b) is a schematic diagram showing the effect of the present invention on estimating the carbon dioxide concentration at the outlet of the absorption column when noise interference is added and the purified air temperature at the outlet of the absorption column is measured at a failure.
Fig. 5 (a) is a schematic diagram showing the effect of the present invention on estimating the lean liquid load at the inlet of the absorption tower when noise interference is added and the temperature of the rich liquid at the outlet of the absorption tower is measured to be faulty.
Fig. 5 (b) is a schematic diagram showing the effect of the present invention on estimating the carbon dioxide concentration at the outlet of the absorption column when noise interference is added and the temperature of the rich liquid at the outlet of the absorption column is measured.
Detailed Description
As shown in fig. 1, a kind of based on spiritChemisorbed CO via network and rolling time domain estimation 2 A trapping system dynamic robust soft measurement system, comprising: rolling time domain estimator (MHE) 1, unscented kalman filter estimator (UKF) 2, BP neural network reduced model 3, chemical absorption CO 2 The system model 4 is captured. A rolling time domain estimator (MHE) 1 estimates a state quantity at a current time based on a state estimation value at a past time and an input quantity; the unscented Kalman filter estimator (UKF) 2 is used to update the covariance matrix of the rolling time domain estimator (MHE) 1; the output of the BP neural network reduced model 3 is used to calculate the cost of arrival Φ in the rolling time domain estimator (MHE) 1 T The method comprises the steps of carrying out a first treatment on the surface of the Chemical absorption of CO 2 The trapping whole system model 4 provides dynamic data for BP neural network identification to obtain a high-precision BP neural network simplified model 3, and chemically absorbs CO 2 A schematic flow diagram of the trapping system model 4 is shown in fig. 2.BP neural network simplified model 3 is chemical absorption CO 2 Capturing lean liquid load at inlet of absorption tower and purified gas CO at outlet of absorption tower at current moment in overall system model 4 2 Concentration, purification gas temperature at the outlet of the absorption tower, rich liquid temperature at the outlet of the absorption tower, lean liquid flow at the inlet of the absorption tower and flue gas CO at the inlet of the absorption tower 2 Concentration, absorption tower inlet flue gas flow, reboiler temperature, absorption tower inlet lean solution load at next moment, absorption tower outlet purified gas CO 2 And (3) using the simplified model of the concentration, the purified gas temperature at the outlet of the absorption tower and the rich liquid temperature at the outlet of the absorption tower as a state updating equation f (x) of the rolling time domain estimator 1. The rolling time domain estimator 1 outputs the BP neural network simplified model 3 as the arrival cost phi T Hypothesis value { x for state quantity t0 -a }; the rolling time domain estimator 1 accepts the output { x } from the BP neural network reduced model 3 t0+1 -for calculating the arrival cost Φ T Obtaining the estimated value of the state quantity at the next moment by minimizing the arrival costThe unscented Kalman filter estimator 2 rolls the estimated value of the state quantity of the time domain estimator 1 according to the corresponding time instant +.>And error covariance matrix pi T-N-1 To calculate and update the error covariance matrix pi of the rolling time domain estimator 1 at the next moment T-N 。
Correspondingly, a chemical absorption CO based on neural network and rolling time domain estimation 2 The dynamic robust soft measurement method of the trapping system comprises the following steps:
(1) Selecting lean liquid load at inlet of absorption tower and purifying gas CO at outlet of absorption tower 2 The concentration is the to-be-estimated value, the temperature of purified gas at the outlet of the absorption tower and the temperature of rich liquid at the outlet of the absorption tower are selected as the measurement values, and the flow of lean liquid at the inlet of the absorption tower and the CO of flue gas at the inlet of the absorption tower are selected 2 The concentration, the flue gas flow at the inlet of the absorption tower and the temperature of the reboiler are used as input quantities, and initial values are set;
(2) Setting the time domain window size N, the state covariance Q, the measurement covariance R and the state error covariance matrix pi in the rolling time domain estimator 0 And state quantity constraint values and prior distribution factors beta of state covariance Qu, measurement covariance Ru, scale factors alpha, scale parameters k and x in the unscented Kalman filter estimator;
(3) Assuming that the state variable x in the BP neural network simplified model 3 contains gaussian white noise W (k), the observed value y contains gaussian white noise V (k), and the system nonlinearity is formed as shown in formula (1):
wherein x is k Represents k moment of lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower 2 Vector of concentration, purification gas temperature at the outlet of the absorption tower and rich liquid temperature at the outlet of the absorption tower. u (u) k Represents k moment of flow of lean liquid at the inlet of the absorption tower and CO at the inlet of the absorption tower 2 The vector of concentration, absorber inlet flue gas flow and reboiler temperature. w (w) k And v k The state noise and the measurement noise at the k time are respectively. f (x, u) is a simplified system model obtained by the BP neural network. Due to the temperature and suction of the purified gas at the outlet of the absorption towerThe temperature of the rich liquid at the outlet of the receiving tower is a measurable quantity and is used as a state quantity, so L is linear transformation with the diagonal angle of 1.
(4) When T is less than or equal to N, obtaining a state estimation value of the current moment of the rolling time domain estimator 1 by solving an objective function for minimizing full information estimation, such as a formula (2), wherein the state error covariance matrix defaults to pi 0 ：
(5) When T > N, the state estimation value of the current moment of the rolling time domain estimator 1 is obtained by solving the objective function Φ of the minimum approximation estimation as in formula (3):
(6) The unscented Kalman filter estimator 2 uses UT conversion from the state estimate at time T-N-1Sum error covariance matrix pi T-N-1 In this process, the state estimate at time T-N-1 is recorded +.>Is->Error covariance matrix pi at time T-N-1 T-N-1 Is P xp Generating a simga point set { χ ] by adopting a symmetrical sampling method i -as in equation (4):
where n is the state dimension, taken here as 4, k=2n+1 is the number of sigma sampling points, W i (i=0, 1 … … K-1) is a sigma point setThe weight value of the weight value is calculated,and->The weights of their mean and covariance, respectively. λ=α 2 (n+k) -n is used to adjust sigma point to +.>Is a distance of (3).
(7) Calculating sigma sampling point predicted state point of BP neural network simplified model 3As in formula (5):
(8) Sampling points according to predicted statesAnd corresponding weight W i Calculating the predicted state mean->Sum covariance P k+1|k As in formula (6):
(9) Sampling points according to predicted statesCalculating a measurement sampling point of the BP neural network simplified model 3, as shown in a formula (7):
(10) Calculating the mean and covariance of the predicted measurements as in equation (8):
(11) Calculating and updating the error covariance matrix pi of the next instant of the rolling horizon estimator 1 T-N As in formula (9):
P y to observe the variance matrix, P xy Is the covariance matrix of the observations and the state vector.
(12) In the following period, steps (5) to (11) are repeatedly executed, the unscented Kalman filter estimator 2 is used for updating to obtain an error covariance matrix of a new moment, and then the rolling time domain estimator 1 is used for solving the state estimation value of the next moment.
The technical effects of the invention are further illustrated in detail below with a specific measurement example:
(1) The input, measured and estimated quantities in the chemisorbed carbon capture system were determined as shown in table 1:
TABLE 1 input, measured and estimated quantities in a chemisorbed carbon capture system
| Input quantity | Measurement quantity | Estimated quantity |
| Lean liquid flow u at inlet of absorption tower 1 | Temperature y of purified gas at outlet of absorption tower 1 | Lean liquid load a at inlet of absorption tower 1 |
| Absorption tower inlet flue gas CO 2 Concentration u 2 | Rich liquid temperature y at outlet of absorption tower 2 | Purifying gas CO at outlet of absorption tower 2 Concentration a 2 |
| Absorption tower inlet flue gas flow u 3 | ||
| Reboiler temperature u 4 |
(2) Setting sampling time T=30s, inputting the estimated quantity and the measured quantity into a neural network by using input quantity data, outputting the estimated quantity and the measured quantity into the neural network, and establishing a core parameter simplified model of the chemical absorption carbon capture system by using a BP neural network tool box. The neural network comprises a hidden layer, the number of neurons is 10, and the training function is traingdm;
(3) Setting a simulated Gaussian white noise variance: q0=1×diag ([ 0.001 0.00001 0.0002 0.001], 0)
Setting UKF estimator related parameters as shown in a formula (10):
setting MHE estimator-related parameters as shown in equation (11):
(4) Selecting a state initial point:
absorption tower outlet purge gas temperature= 328.3493K
Lean liquid load at absorber outlet= 0.2944
Absorption tower outlet carbon dioxide concentration=0.0879
Absorption tower outlet rich liquid temperature= 317.5032K
Absorption tower inlet lean liquid flow = 457.0438
Absorption column inlet carbon dioxide concentration = 0.1927
Absorption tower inlet flue gas flow = 573.2017
Reboiler temperature= 389.9804K
(5) In order to simulate the faults of the purified air temperature measuring instrument at the outlet of the absorption tower under the actual condition, a fault superposition signal with the constant magnitude of 0.1 is added to the purified air temperature measuring signal at t=2250s.
(6) In order to simulate the faults of the absorption tower outlet rich liquid temperature measuring instrument under the actual condition, a fault superposition signal with the magnitude linearly increasing along with the time and keeping 0.1 unchanged after reaching 0.1 is added to the rich liquid temperature measuring signal at t=3000 s.
The dynamic robust soft measurement effect based on the neural network and the rolling time domain estimation is shown in the accompanying figures 3 (a) -5 (b). For convenient observation and comparison, the sampling period is 30 seconds for taking points and drawing. Simulation results show that the chemical absorption CO is estimated based on a neural network and a rolling time domain 2 The dynamic robust soft measurement method of the trapping system has the advantages of small estimation error and good estimation effect.
The invention absorbs CO chemically 2 A trapping system, which uses BP neural network to obtain a simplified model of the core parameters therein,selecting lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower by using a rolling time domain estimation algorithm 2 The concentration is the to-be-estimated value, the temperature of purified gas at the outlet of the absorption tower and the temperature of rich liquid at the outlet of the absorption tower are selected as the measurement values, and the flow of lean liquid at the inlet of the absorption tower and the CO of flue gas at the inlet of the absorption tower are selected 2 The concentration, the flow of flue gas at the inlet of the absorption tower and the temperature of the reboiler are input. On one hand, the carbon dioxide concentration at the outlet of the absorption tower and the lean liquid load at the inlet of the absorption tower can be accurately estimated; in addition, by selecting the combination mode of MHE and UKF, the chemical absorption of CO can be realized under the condition of complex operation such as unknown interference, fault and the like 2 Robust estimation of key parameters in a trapping system.
According to the invention, the carbon dioxide concentration at the outlet of the absorption tower and the lean liquid load at the inlet of the absorption tower can be accurately estimated by using a neural network and a rolling time domain estimation algorithm; at the same time, can absorb CO chemically under the complex operation conditions of noise, unknown interference, faults and the like 2 The concentration of carbon dioxide at the outlet of the absorption tower and the load of lean liquid at the inlet of the absorption tower in the capturing system provide accurate estimation and guidance for system operation control.
Claims (4)
1. Chemical absorption CO based on neural network and rolling time domain estimation 2 The dynamic robust soft measurement method of the trapping system is characterized by comprising the following steps of:
step (1), selecting lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower 2 The concentration is the to-be-estimated value, the temperature of purified gas at the outlet of the absorption tower and the temperature of rich liquid at the outlet of the absorption tower are selected as the measurement values, and the flow of lean liquid at the inlet of the absorption tower and the CO of flue gas at the inlet of the absorption tower are selected 2 The concentration, the flue gas flow at the inlet of the absorption tower and the temperature of the reboiler are used as input quantities, and initial values are set;
step (2), setting the time domain window size N, the state covariance Q, the measurement covariance R and the state error covariance matrix pi in the rolling time domain estimator (1) 0 And state quantity constraint values and prior distribution factors beta of state covariance Qu, measurement covariance Ru, scale factors alpha, scale parameters k and x in the unscented Kalman filter estimator (2);
step (3), assuming that the state variable x in the BP neural network simplified model (3) contains Gaussian white noise W (k), and the observed value y contains Gaussian white noise V (k), forming system nonlinearity, such as formula (1):
wherein χ is k Represents k moment of lean liquid load at the inlet of the absorption tower and purified gas CO at the outlet of the absorption tower 2 Vector of concentration, purification gas temperature at the outlet of the absorption tower and rich liquid temperature at the outlet of the absorption tower; u (u) k Represents k moment of flow of lean liquid at the inlet of the absorption tower and CO at the inlet of the absorption tower 2 Vector of concentration, absorption tower inlet flue gas flow and reboiler temperature; w (w) k And v k The state noise and the measurement noise at the moment k are respectively; f (x, u) is a BP neural network simplified model (3); because the temperature of the purified gas at the outlet of the absorption tower and the temperature of the rich liquid at the outlet of the absorption tower are measurable and serve as state quantity, L is linear transformation with the diagonal angle of 1;
step (4), when T is less than or equal to N, obtaining a state estimation value of the current moment of the rolling time domain estimator (1) through solving an objective function phi for minimizing full information estimation, such as a formula (2), and pi 0 The initial matrix is the state error covariance:
step (5), when T > N, obtaining a state estimation value of the current moment of the rolling time domain estimator (1) by solving an objective function Φ of minimizing approximate estimation, as shown in formula (3):
step (6), using UT conversion by unscented Kalman filter estimator (2), estimating value from state of T-N-1 timeSum error covariance matrix pi T-N-1 In this process, the state estimate at time T-N-1 is recorded +.>Is->Error covariance matrix pi at time T-N-1 T-N-1 Is P xp A symmetrical sampling method is adopted to generate a sigma point set { χ } i -as in equation (4):
where n is the state dimension, k=2n+1 is the number of sigma sampling points, W i (i=0, 1 … … K-1) is the weight of the sigma point set,and->The weights of the mean and covariance, respectively; λ=α 2 (n+k) -n is used to adjust sigma point to +.>Is a distance of (2);
step (7), calculating a prediction state sampling point of sigma sampling points of the BP neural network simplified model (3)As in formula (5):
step (8), sampling points according to the prediction stateAnd corresponding weight W i Calculating the predicted state mean->Sum covariance P k+1|k As in formula (6):
step (9), sampling points according to the prediction stateCalculating measurement sampling points of the BP neural network simplified model (3), such as a formula (7):
step (10), calculating the mean and covariance of the predicted measurements, as in equation (8):
P y to observe the variance matrix, P xy Is the covariance matrix of the observations and the state vector;
step (11), calculating and updating an error covariance matrix pi of the next moment of the rolling time domain estimator (1) T-N As in formula (9):
and (12) repeatedly executing the steps (5) to (11) in the following period, updating to obtain an error covariance matrix of the new moment, and then solving a state estimation value of the next moment by using the rolling time domain estimator (1).
2. The method of claim 1, wherein in step (2), the scale factor α satisfies 10 -4 ≤α<1, a step of; the value of the scale parameter k is 0; the prior distribution factor beta of x takes a value of 2 for Gaussian distribution; the state dimension n takes a value of 4.
3. A neural network and rolling horizon estimation based chemisorption CO implementing the method of claim 1 2 A trapping system dynamic robust soft measurement system, comprising: rolling time domain estimator (1), unscented Kalman filter estimator (2), BP neural network simplified model (3), chemical absorption CO 2 A trapping system model (4); the BP neural network simplified model (3) is chemical absorption CO 2 Capturing a simplified model among core parameters of the overall system model (4) and updating an equation f (x) as a state of the rolling time domain estimator (1); the output quantity of the rolling time domain estimator (1) to the BP neural network simplified model (3) is the assumed value { x ] of the arrival cost function to the state quantity t0 -a }; the rolling time domain estimator (1) receives the output { x } from the BP neural network reduced model (3) t0+1 -for calculating the arrival cost Φ T Obtaining the estimated value of the state quantity at the next moment by minimizing the arrival costThe unscented Kalman filter estimator (2) rolls the estimated value of the state quantity of the time domain estimator (1) according to the corresponding timeAnd error covariance matrix pi T-N-1 To calculate and update an error covariance matrix pi for the next instant of the rolling horizon estimator (1) T-N 。
4. A system according to claim 3, characterized in that the chemical absorption of CO 2 Capturing the overall system model (4) core parameters includes: lean liquid load at inlet of absorption tower and purified gas CO at outlet of absorption tower at current moment 2 Concentration, temperature of purified gas at outlet of absorption tower and outlet of absorption towerRich liquid temperature, lean liquid flow rate at inlet of absorption tower and flue gas CO at inlet of absorption tower 2 Concentration, absorption tower inlet flue gas flow, reboiler temperature, absorption tower inlet lean solution load at next moment, absorption tower outlet purifier CO 2 Concentration, temperature of purified gas at the outlet of the absorption tower and temperature of rich liquid at the outlet of the absorption tower.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110617183.1A CN113393905B (en) | 2021-06-03 | 2021-06-03 | Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110617183.1A CN113393905B (en) | 2021-06-03 | 2021-06-03 | Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113393905A CN113393905A (en) | 2021-09-14 |
| CN113393905B true CN113393905B (en) | 2024-03-22 |
Family
ID=77619826
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110617183.1A Active CN113393905B (en) | 2021-06-03 | 2021-06-03 | Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113393905B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116679572B (en) * | 2023-08-03 | 2023-09-29 | 北京绿能碳宝科技发展有限公司 | Carbon dioxide trapping self-learning method based on deep Q learning network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763036A (en) * | 2009-12-29 | 2010-06-30 | 江苏大学 | Lysine fermentation process feeding prediction control system and method based on fuzzy neural network |
| CN110026068A (en) * | 2019-04-08 | 2019-07-19 | 东南大学 | A kind of large-scale coal fired power plant CO based on Neural network inverse control2Trapping system and feed forward control method |
| CN110737198A (en) * | 2019-10-09 | 2020-01-31 | 东南大学 | Large-scale coal-fired power plant CO based on BP neural network2Capture system prediction control method |
| CN110764419A (en) * | 2019-11-15 | 2020-02-07 | 江苏方天电力技术有限公司 | CO of large coal-fired power station2Capture global scheduling and predictive control system and method |
| CN111259927A (en) * | 2020-01-08 | 2020-06-09 | 西北工业大学 | Rocket engine fault diagnosis method based on neural network and evidence theory |
| CN112511056A (en) * | 2020-11-09 | 2021-03-16 | 广东电网有限责任公司广州供电局 | Robust generator dynamic state estimation method based on phasor measurement |
-
2021
- 2021-06-03 CN CN202110617183.1A patent/CN113393905B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763036A (en) * | 2009-12-29 | 2010-06-30 | 江苏大学 | Lysine fermentation process feeding prediction control system and method based on fuzzy neural network |
| CN110026068A (en) * | 2019-04-08 | 2019-07-19 | 东南大学 | A kind of large-scale coal fired power plant CO based on Neural network inverse control2Trapping system and feed forward control method |
| CN110737198A (en) * | 2019-10-09 | 2020-01-31 | 东南大学 | Large-scale coal-fired power plant CO based on BP neural network2Capture system prediction control method |
| CN110764419A (en) * | 2019-11-15 | 2020-02-07 | 江苏方天电力技术有限公司 | CO of large coal-fired power station2Capture global scheduling and predictive control system and method |
| CN111259927A (en) * | 2020-01-08 | 2020-06-09 | 西北工业大学 | Rocket engine fault diagnosis method based on neural network and evidence theory |
| CN112511056A (en) * | 2020-11-09 | 2021-03-16 | 广东电网有限责任公司广州供电局 | Robust generator dynamic state estimation method based on phasor measurement |
Non-Patent Citations (2)
| Title |
|---|
| 噪声方差不确定约束系统的滚动时域估计;赵海艳;陈虹;;控制与决策;20080215(第02期);全文 * |
| 基于流程模拟与Elman神经网络的燃煤电厂CO_2吸收系统动态数据检验方法;梁海文;周栋;;华东电力;20100724(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113393905A (en) | 2021-09-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106920007B (en) | PM based on second-order self-organizing fuzzy neural network2.5Intelligent prediction method | |
| CN105784556B (en) | A kind of air fine particles PM based on Self-organized Fuzzy Neural Network2.5Flexible measurement method | |
| Fisher et al. | On the equivalence between Kalman smoothing and weak‐constraint four‐dimensional variational data assimilation | |
| CN105093122B (en) | Emergency light battery SOC method of estimation based on the adaptive SQKF of strong tracking | |
| CN107290965B (en) | Adaptive soft-sensor prediction technique based on local weighted Bayesian network | |
| CN104914723B (en) | Industrial process soft-measuring modeling method based on coorinated training partial least square model | |
| CN113393905B (en) | Chemical absorption CO 2 Dynamic robust soft measurement system and method for trapping system | |
| CN109088407B (en) | Power distribution network state estimation method based on deep belief network pseudo-measurement modeling | |
| CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
| CN105588930A (en) | Method for measuring parameters of soil in greenhouse | |
| Li et al. | Modeling the corrosion rate of carbon steel in carbonated mixtures of MDEA-based solutions using artificial neural network | |
| CN103645635A (en) | A ship motion controller based on a simulation annealing-reinforcement learning algorithm | |
| CN104283529B (en) | High order volume Kalman filtering method for square root with unknown measurement noise variance | |
| Lim | An estimation model of fine dust concentration using meteorological environment data and machine learning | |
| Wang et al. | Robust monitoring of solvent based carbon capture process using deep learning network based moving horizon estimation | |
| CN111915006B (en) | Tail gas hydrogen sulfide concentration real-time monitoring method based on supervised SAE | |
| CN111707945A (en) | Method for observing internal state of proton exchange membrane fuel cell | |
| Loganathan et al. | Estimation of air quality index using multiple linear regression | |
| Lin et al. | Multi-sensor information fusion extended Kalman particle filter | |
| Li et al. | Knowledge-based genetic algorithms data fusion and its application in mine mixed-gas detection. | |
| Chen et al. | Modeling of wastewater treatment process using recurrent neural network | |
| CN112598178A (en) | Atmospheric pollutant concentration prediction method based on Bayesian framework | |
| Wahab et al. | A Review on Data Reconciliation and Gross Error Detection for Process Plant Energy Management | |
| Ng et al. | Sparse matrix computation for air quality forecast data assimilation | |
| Qiao et al. | Fault diagnosis of slurry pH data base on autoregressive integrated moving average and least squares support vector machines |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |