CN107908108B - Natural gas absorption tower desulfurization process control method based on UKF and GDHP - Google Patents

Natural gas absorption tower desulfurization process control method based on UKF and GDHP Download PDF

Info

Publication number
CN107908108B
CN107908108B CN201711115675.0A CN201711115675A CN107908108B CN 107908108 B CN107908108 B CN 107908108B CN 201711115675 A CN201711115675 A CN 201711115675A CN 107908108 B CN107908108 B CN 107908108B
Authority
CN
China
Prior art keywords
control
desulfurization process
natural gas
absorption tower
network
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
Application number
CN201711115675.0A
Other languages
Chinese (zh)
Other versions
CN107908108A (en
Inventor
甘丽群
刘华超
周伟
汪波
李晓亮
易军
李太福
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Science and Technology
Original Assignee
Chongqing University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing University of Science and Technology filed Critical Chongqing University of Science and Technology
Priority to CN201711115675.0A priority Critical patent/CN107908108B/en
Publication of CN107908108A publication Critical patent/CN107908108A/en
Application granted granted Critical
Publication of CN107908108B publication Critical patent/CN107908108B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
    • G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only

Abstract

The invention provides a natural gas absorption tower desulfurization process control method based on UKF and GDHP. And modeling the desulfurization process of the natural gas absorption tower by using a BP neural network, carrying out a desulfurization process control simulation experiment by using the model as a controlled object, and continuously updating the optimized weight according to the control error and the performance index function until an optimal control signal is obtained, thereby realizing the optimal control of the desulfurization process. The desulfurization process of the natural gas absorption tower is complex, the characteristics of uncertainty, nonlinearity, strong coupling, dynamic property and the like are presented, an accurate mathematical model is difficult to establish, and the control difficulty is high. The natural gas absorption tower desulfurization process control method based on the UKF and the GDHP is provided aiming at the problems of low control precision, large time lag, instability and the like of the existing desulfurization process control method, so that the stability and the control precision of a control system are ensured, the response time is reduced, and the real-time accurate control of the desulfurization process is really realized.

Description

Natural gas absorption tower desulfurization process control method based on UKF and GDHP
Technical Field
The invention relates to a natural gas absorption tower desulfurization process control technology, in particular to a natural gas absorption tower desulfurization process control method based on combination of Unscented Kalman Filtering (UKF) and global quadratic heuristic dynamic programming (GDHP).
Background
The natural gas is used as a high-quality and clean energy source and chemical raw material, is convenient to use and has higher comprehensive economic benefit. China has abundant natural gas resources, but about 30 percent of natural gas contains a large amount of sulfur elements, wherein H2The natural gas reserves with an S content of more than 1% account for 1/4 of the total reserves. H2The existence of S not only causes the corrosion of equipment and pipelines and harms human health, but also causes the pollution of the environment by combustion products. Thus, in the desulfurization of natural gas, H2S content control is particularly important.
The natural gas desulfurization absorption tower is an important component of a natural gas purification device and directly influences the natural gas purification effect. The natural gas raw material gas enters an absorption tower to be fully contacted with a Methyldiethanolamine (MDEA) solution in the tower for reaction, so that the aim of desulfurization is fulfilled, a physical and chemical reaction and a phase reaction are simultaneously performed in the whole process, the material conversion and energy transfer are involved, the characteristics of uncertainty, nonlinearity, strong coupling property, dynamic property and the like are greatly influenced by various uncertain factors, an accurate mathematical model is difficult to establish, and great difficulty is brought to the control of the desulfurization process of the absorption tower.
The existing control technology is mostly PID single loop control or simple cascade control, the automation degree of a control system is not high, and too much control parameters are adjusted by depending on expert experience, so that the control system has larger hysteresis, lower control precision, and difficult guarantee of the stability of the control system, and is difficult to achieve real-time accurate control.
Disclosure of Invention
The utility model provides a natural gas absorption tower desulfurization process control method based on UKF and GDHP to solve the control accuracy that exists in the present natural gas absorption tower desulfurization process control technique and hang down, the time lag is big, and problems such as control system unstability guarantee the natural gas desulfurization effect.
In order to solve the technical problems, the application adopts the following technical scheme:
a natural gas absorption tower desulfurization process control method based on UKF and GDHP is characterized by comprising the following steps:
step 1: by analyzing the desulfurization process of the absorption tower, determining that the main factors influencing the desulfurization effect of the natural gas are the acid natural gas treatment amount and the alcohol amine solution circulation amount which are respectively expressed by u1 and u2, and thus forming a control variable u ═ u1, u 2;
step 2: determining sample data output by the input sample data of the desulfurization process model, and establishing a natural gas absorption tower desulfurization process model by adopting a BP neural network;
and step 3: setting a desired control target value
Figure GDA0002765264250000021
Updating the evaluation network and the execution network weight in the GDHP control method by using a UKF algorithm, and respectively obtaining control signals u (k) ═ u1 and u2 through the execution network and the evaluation network]And cost function J (k) and its partial derivatives of system state x (k)
Figure GDA0002765264250000022
Establishing a natural gas absorption tower desulfurization process control method of UKF-GDHP;
and 4, step 4: inputting the control signal u (k) ═ u1, u2 and the current time system state x (k) ═ x1, x2 as the model of the desulfurization process of the absorption tower, thereby obtaining a system output x (k + 1);
and 5: and (e) (k) calculating a control error E (k), if the control error is smaller than the expected error, finishing the training, otherwise, returning to the step 3.
As a further illustration, the step 3 is specifically performed according to the following steps:
step 3-1: updating the evaluation network and the execution network weight by adopting a UKF algorithm according to the control error E (k);
step 3-2: calculating a control signal u (k);
step 3-3: the evaluation network outputs J (k +1) and λ (k +1) are calculated.
For further explanation, the process of updating the weight by using the UKF algorithm in the step 3-1 is as follows:
(1) initializing system parameters;
(2) calculating a Sigma point state vector;
(3): performing one-step prediction of system state and covariance matrix;
(4): calculating system observation and covariance matrixes;
(5): calculating a Kalman gain;
(6): and updating a system state estimation matrix and a covariance matrix.
For further explanation, in step 5, the control error e (k) is calculated by the formula:
E(k)＝α||E1||+(1-α||E2||)
wherein the content of the first and second substances,
Figure GDA0002765264250000031
Figure GDA0002765264250000032
the function u (k) is a utility function, which is a function of the control signal u (k) and the cost function j (k).
Compared with the prior art, the technical scheme that this application provided, the technological effect or advantage that have are: in the process of desulfurization of the absorption tower, the method has high control precision and high convergence rate, can improve the stability and control precision of a control system, reduce the response time of the control system and ensure the desulfurization effect of natural gas.
Drawings
FIG. 1 is a functional block diagram of the present invention;
FIG. 2 is a schematic diagram of a natural gas absorption tower desulfurization process model;
FIGS. 3-6 are schematic diagrams of model test results of a desulfurization process of a natural gas absorption tower;
FIG. 3H2A schematic diagram of S content prediction;
FIG. 4H2S content prediction relative error schematic diagram;
FIG. 5 CO2Content prediction schematic diagram;
FIG. 6 CO2Content prediction relative error schematic diagram;
FIG. 7 is a schematic view of the UKF-ADHDP control structure;
FIGS. 8-11 are schematic diagrams of control curves for a desulfurization process in a natural gas absorber;
FIG. 8 Natural gas purge H2S content control curve diagram;
FIG. 9 CO in Natural gas purge2A content control curve diagram;
FIG. 10 is a schematic view of a control curve for natural gas feed gas treatment;
FIG. 11 is a schematic view of a control curve of the circulation amount of an alcohol amine solution.
Detailed Description
The application provides a natural gas absorption tower desulfurization process control method based on UKFGDHP, and the invention principle block diagram is shown in figure 1. Referring to the prior art means, the technical scheme provided by the application has the following technical effects or advantages: the method adopts an intelligent algorithm for controlling the desulfurization process of the absorption tower, has higher control precision, can reduce the response time of a control system, can automatically adjust control parameters in real time, improves the stability of the control system, and really achieves the aim of real-time control.
In order to better understand the technical solutions, the technical solutions will be described in detail below with reference to fig. 2 to 11 of the specification and specific embodiments.
Firstly, entering the step 1: and selecting two parameters of the acid natural gas treatment amount and the circulation amount of the alcohol amine solution for absorbing the acid gas to form a control variable u ═ u1, u 2.
Step 2: using BP neural network to input1～inputnAnd x1～xnAnd training and checking the sample as an input and output sample so as to establish an absorption tower desulfurization process model. Wherein input is [ x1, x2, u1, u2]，x＝[x1，x2]N represents the number of samples, u1 and u2 represent the handling capacity of the raw natural gas and the circulation capacity of the alcohol amine solution in unit time respectively, and x1 and x2 represent H in the natural gas purified gas respectively2S content (mg/m)3) And CO2Content (%).
In this example, a desulfurization process model of the absorption tower as shown in FIG. 2 was establishedThe number of neurons in the input layer is 4, and the number of neurons in the output layer is 2; empirically, the hidden layer node is selected as
Figure GDA0002765264250000041
(x is an input layer node, y is an output layer node, and a is 1,2, … 9), and selecting an implicit layer node with the highest modeling test precision as 10 through sequential experiments; the hidden layer transfer function is a tansig function, and the output layer transfer function is a purelin function; the minimum value of the expected error is 0.0001, and the learning efficiency of the correction weight is 0.05. The modeling sample data is the actual production data of the plain gas field, the total number of the modeling sample data is 500 groups, 80% of the sample data is randomly selected to be used for model training, and the rest 20% of the sample data is used for model testing.
Setting the input of an absorption tower desulfurization process model as P, the number of input neurons as r, the number of neurons in a hidden layer as s1, a corresponding activation function as h1 and the output of the hidden layer as a 1; the number of neurons in the output layer is s2, the corresponding activation function is h2, the output is a2, and the target vector is T.
The step 2 of establishing the absorption tower desulfurization process model specifically comprises the following steps:
step 2-1: initializing, setting an initial value of iteration times g as 0, and simultaneously giving a random value in a (0,1) interval of W1, W2, B1 and B2;
step 2-2: random input sample Pj；
Step 2-3: for input sample PjForward computing the input and output of each layer of neuron of BP neural network;
the output of the ith neuron of the hidden layer is:
Figure GDA0002765264250000051
the output of the s-th neuron of the output layer is:
Figure GDA0002765264250000052
step 2-4: calculating an error e (g) based on the desired output T and the actual output a2 (g);
the error function is defined as:
Figure GDA0002765264250000053
step 2-5: judging whether the error E (g) meets the requirement, if not, entering the step 2-6, and if so, entering the step 2-8;
step 2-6: judging whether the iteration times g +1 are larger than the maximum iteration times, if so, entering the step 2-8, otherwise, entering the step 2-7;
step 2-7: and calculating the weight correction quantity delta W and correcting the weight.
Output layer weight change
For the weight from the ith input to the kth output, there are:
Figure GDA0002765264250000061
wherein the content of the first and second substances,ki＝(tk-a2k)·h2′＝ek·h2′，ek＝tk-a2k。
Figure GDA0002765264250000062
② hidden layer weight change
For the weight from the jth input to the ith output, there are:
Figure GDA0002765264250000063
wherein the content of the first and second substances,
Figure GDA0002765264250000064
the same can be obtained:
Δb1i＝η·ij
in the formula, eta is learning efficiency; making g equal to g +1, and skipping to the step 3;
step 2-8: judging whether all training samples are finished, if so, finishing modeling, and otherwise, continuing to jump to the step 2-2;
through the above process, the prediction effect of the BP neural network can be obtained as shown in fig. 3 and 5, and the corresponding prediction error is shown in fig. 4 and 6. 3-6, the BP neural network training has higher precision in establishing the absorption tower desulfurization process model, can accurately predict the output of the system, and lays a foundation for the research of the natural gas desulfurization process control method.
And step 3: setting a desired control target value
Figure GDA0002765264250000071
Updating the evaluation network and the execution network weight in the GDHP control method by using a UKF algorithm, and respectively obtaining control signals u (k) ═ u1 and u2 through the execution network and the evaluation network]And cost function J (k) and its partial derivatives of system state x (k)
Figure GDA0002765264250000072
The method for controlling the desulfurization process of the natural gas absorption tower by establishing the UKF-GDHP has the control structure shown in figure 7: the Action-UKF is an execution network, and the input and output are respectively a system state x (k) and a control signal u (k); the Controlled Object is a model network, the input is a system state x (k) and a control signal u (k), and the output is a next time state x (k +1) of the system; Critic-UKF is an evaluation network, the inputs are x (k +1) and u (k +1), and the output is a performance index function J (k +1) and partial derivatives thereof to the system state
Figure GDA0002765264250000073
The training of the execution network and the evaluation network respectively aims at minimizing a control error E (k), a performance index function J (k) and a partial derivative function lambda (k) of the system state x (k), and a dotted line represents a network weight adjustment path.
The UKF algorithm sampling point and the corresponding weight calculation formula are as follows:
(1) calculating 2n +1 sigma points
Figure GDA0002765264250000074
Figure GDA0002765264250000075
Figure GDA0002765264250000076
(2) Calculating the corresponding weight of the sampling point
Figure GDA0002765264250000077
Figure GDA0002765264250000078
Figure GDA0002765264250000079
Parameter λ ═ α2The (n + tau) -n is a scaling parameter, the selection of alpha controls the distribution state of the sampling points, tau is a parameter to be selected, and the value of tau is not particularly limited, but usually the matrix (n + lambda) P is ensured to be a semi-positive definite matrix. The candidate parameter sigma is a non-negative weight coefficient which can combine the dynamic difference of high-order terms in the equation, so that the influence of the high-order terms can be included, and the error is reduced.
Figure GDA0002765264250000081
The ith row or column of the matrix root mean square,
Figure GDA0002765264250000082
the weights of the matrix and covariance are respectively.
Step 3-1: weight value updating
(1) Initializing system parameters;
(2) calculating a Sigma point state vector;
(3) performing one-step prediction of system state and covariance matrix;
(4) calculating system observation and covariance matrixes;
(5) calculating a Kalman gain;
(6) updating the system state estimation matrix and the covariance matrix as follows:
Figure GDA0002765264250000083
in the formula (I), the compound is shown in the specification,
Figure GDA0002765264250000084
the matrix is estimated for the system state at time k,
Figure GDA0002765264250000085
g (k +1| k) is the system observation matrix at time k,
Figure GDA0002765264250000086
a prediction matrix is observed for the system at time k;
Figure GDA0002765264250000087
in the formula (I), the compound is shown in the specification,
Figure GDA0002765264250000088
a matrix covariance matrix is estimated for the system at time k,
Figure GDA0002765264250000089
a covariance matrix of an observation matrix of the system at the moment k;
step 3-2: computing control signals u (k)
Will execute the network hidden layer weight Wa1And the output layer weight Wa2Two sets of parameters as state vectors for the execution networkXa。
Figure GDA00027652642500000810
XaAnd (3) continuously updating and optimizing through the step 3-1 to finally obtain a group of optimal weights.
The output of the network hidden layer after updating is executed as follows:
La＝x(k)*Wa1
where x (k) is the system state at time k, i.e., the control output.
The updated execution network output layer outputs:
u(k)＝Qa*Wa2
in the formula, Qa＝(1-exp(-La))/(1+exp(-La) U (k) is the control signal output by the execution network.
Step 3-3: computing evaluation network output J (k +1) and lambda (k +1)
Will evaluate the weight W of the hidden layer of the networkc1And the output layer weight Wc2Two groups of parameters are used as state vector X of evaluation networkc：
Figure GDA0002765264250000091
XcAnd (3) continuously updating and optimizing through the step 3-1 to finally obtain a group of optimal weights, and evaluating the output of the network hidden layer after updating as follows:
Lc＝input*WC1
wherein, input ═ x (k +1), u (k +1) ], and the output of the evaluation network output layer after updating is:
J(k+1)＝Qc*WC2
Figure GDA0002765264250000092
wherein Q isc＝(1-exp(-Lc))/(1+exp(-Lc))，QcThe expressed function is a sigmod function, which is used as a threshold function for neural networks.
And 4, step 4: and (3) inputting the control signal u (k) ═ u1, u2 and the system state x (k) ═ x1, x2 at the current time as the desulfurization process model of the absorption tower, thereby obtaining a system output x (k + 1).
And 5: and (e) (k) calculating a control error E (k), if the control error is smaller than the expected error, finishing the training, otherwise, returning to the step 3.
In this embodiment, the target values x1 and x2 are set to 1.53 and 1.6 respectively, and through the above process, H in the natural gas purified gas can be obtained2S and CO2The content control curve is shown in figures 8 and 9, and the raw material natural gas treatment capacity and the alcohol amine solution circulation control curve is shown in figures 10 and 11. By analyzing 8-11, the UKF-GDHP method can control the desulfurization process of the absorption tower to meet the actual process requirements, and has the characteristics of high convergence rate, high control precision and the like.
The invention provides a method for controlling a desulfurization process of an absorption tower based on UKF and GDHP. Firstly, training the actual production data of the absorption tower desulfurization by using a BP neural network, and establishing an absorption tower desulfurization process model, thereby bypassing the detailed problem in the desulfurization process mechanism, solving the problem of difficult modeling caused by the complicated desulfurization process, and laying a foundation for the research of the natural gas desulfurization process control method. And then, the established model is used as a controlled object for predicting the control output of the system, the GDHP method is adopted for controlling the desulfurization process of the absorption tower, the UKF algorithm is adopted for updating and optimizing the GDHP evaluation network and the execution network weight, and the UKF-GDHP-based absorption tower desulfurization process control method is established. The method gets rid of the excessive dependence on expert experience for a long time, solves the problems of low control precision, large time lag, unstable control system and the like in the existing absorption tower desulfurization process control technology, really achieves the purpose of real-time and accurate control of the absorption tower desulfurization process, also provides a new idea for solving the similar industrial control problem, and embodies the powerful functions of an artificial intelligence algorithm in the industry.
It should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make variations, modifications, additions or substitutions within the spirit and scope of the present invention.

Claims (2)

1. A natural gas absorption tower desulfurization process control method based on UKF and GDHP is characterized by comprising the following steps:
step 1: by analyzing the desulfurization process of the natural gas absorption tower, determining that the main factors influencing the desulfurization effect of the natural gas are the acid natural gas treatment amount and the alcohol amine solution circulation amount which are respectively expressed by u1 and u2, and thus forming control variables u ═ u1 and u 2;
step 2: determining the input sample data and the output sample data of the desulfurization process model, and establishing a natural gas absorption tower desulfurization process model by adopting a BP neural network, wherein the number of neurons of an input layer is 4, the number of nodes of a hidden layer is 10, the number of neurons of an output layer is 2, a transfer function of the hidden layer is a tansig function, a transfer function of the output layer is a purelin function, the minimum value of an expected error is 0.0001, and the learning efficiency of a corrected weight is 0.05;
and step 3: setting a desired control target value
Figure FDA0002765264240000011
Updating the evaluation network and the execution network weight in the GDHP control method by using a UKF algorithm, and respectively obtaining control signals u (k) ═ u1 and u2 through the execution network and the evaluation network]And cost function J (k) and its partial derivatives of system state x (k)
Figure FDA0002765264240000012
The method for establishing the natural gas absorption tower desulfurization process control of the UKF-GDHP specifically comprises the following steps:
step 3-1: updating the evaluation network and the execution network weight by adopting a UKF algorithm, wherein a system state estimation matrix and a covariance matrix are updated as follows:
Figure FDA0002765264240000013
wherein the content of the first and second substances,
Figure FDA0002765264240000014
the matrix is estimated for the system state at time k,
Figure FDA0002765264240000015
g (k +1| k) is the system observation matrix at time k,
Figure FDA0002765264240000016
a prediction matrix is observed for the system at time k;
Figure FDA0002765264240000017
wherein the content of the first and second substances,
Figure FDA0002765264240000021
a matrix covariance matrix is estimated for the system at time k,
Figure FDA0002765264240000022
a covariance matrix of an observation matrix of the system at the moment k;
step 3-2: calculating a control signal u (k):
optimizing execution network hidden layer weight W of UKFa1And the output layer weight Wa2Two sets of parameters as state vector X for the execution networka：
Figure FDA0002765264240000023
The output of the network hidden layer after updating is executed as follows:
La＝x(k)*Wa1
wherein x (k) is the system state at the moment k, namely the control output;
the updated execution network output layer outputs:
u(k)＝Qa*Wa2
wherein Q isa＝(1-exp(-La))/(1+exp(-La))；
Step 3-3: calculating and evaluating network outputs J (k +1) and lambda (k + 1):
evaluating network hidden layer weight W after UKF optimizationc1And the output layer weight Wc2Two groups of parameters are used as state vector X of evaluation networkc：
Figure FDA0002765264240000024
The output of the evaluation network hidden layer after updating is as follows:
Lc＝input*WC1
wherein, input ═ x (k +1), u (k +1) ], and the output of the evaluation network output layer after updating is:
J(k+1)＝Qc*WC2
Figure FDA0002765264240000025
wherein Q isc＝(1-exp(-Lc))/(1+exp(-Lc))，QcIs a sigmod function used as a threshold function of the neural network;
and 4, step 4: inputting the control signal u (k) ═ u1, u2 and the current time system state x (k) ═ x1, x2 as the model of the desulfurization process of the absorption tower, thereby obtaining a system output x (k + 1);
and 5: and (e) (k) calculating a control error E (k), if the control error is smaller than the expected error, finishing the training, otherwise, returning to the step 3.
2. The UKF and GDHP-based natural gas absorption tower desulfurization process control method according to claim 1, wherein:
in step 5, the control error E (k) is calculated by the formula:
E(k)＝α||E1||+(1-α||E2||)
wherein the content of the first and second substances,
Figure FDA0002765264240000031
Figure FDA0002765264240000032
the function u (k) is a utility function, which is a function of the control signal u (k) and the cost function j (k).
CN201711115675.0A 2017-11-13 2017-11-13 Natural gas absorption tower desulfurization process control method based on UKF and GDHP Active CN107908108B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711115675.0A CN107908108B (en) 2017-11-13 2017-11-13 Natural gas absorption tower desulfurization process control method based on UKF and GDHP

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711115675.0A CN107908108B (en) 2017-11-13 2017-11-13 Natural gas absorption tower desulfurization process control method based on UKF and GDHP

Publications (2)

Publication Number Publication Date
CN107908108A CN107908108A (en) 2018-04-13
CN107908108B true CN107908108B (en) 2021-01-01

Family

ID=61845222

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711115675.0A Active CN107908108B (en) 2017-11-13 2017-11-13 Natural gas absorption tower desulfurization process control method based on UKF and GDHP

Country Status (1)

Country Link
CN (1) CN107908108B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113204189B (en) * 2020-04-28 2023-05-26 大唐环境产业集团股份有限公司 Desulfurization system control model, establishment method thereof and desulfurization system control method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104656441A (en) * 2014-12-29 2015-05-27 重庆科技学院 Natural gas purification process modeling optimization method based on unscented kalman neural network
CN105045941A (en) * 2015-03-13 2015-11-11 重庆科技学院 Oil pumping unit parameter optimization method based on traceless Kalman filtering
CN106021698A (en) * 2016-05-17 2016-10-12 重庆科技学院 Iterative updating-based UKFNN aluminum electrolysis power consumption model construction method
CN106777866A (en) * 2016-11-14 2017-05-31 重庆科技学院 Technology Modeling and optimization method are purified towards energy-saving high sulfur-containing natural gas
CN106777468A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur content natural gas desulfurization process strong tracking evolutionary Modeling method
CN106777465A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur-containing natural gas purify technique dynamic evolutionary modeling and energy conservation optimizing method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8343336B2 (en) * 2007-10-30 2013-01-01 Saudi Arabian Oil Company Desulfurization of whole crude oil by solvent extraction and hydrotreating

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104656441A (en) * 2014-12-29 2015-05-27 重庆科技学院 Natural gas purification process modeling optimization method based on unscented kalman neural network
CN105045941A (en) * 2015-03-13 2015-11-11 重庆科技学院 Oil pumping unit parameter optimization method based on traceless Kalman filtering
CN106021698A (en) * 2016-05-17 2016-10-12 重庆科技学院 Iterative updating-based UKFNN aluminum electrolysis power consumption model construction method
CN106777866A (en) * 2016-11-14 2017-05-31 重庆科技学院 Technology Modeling and optimization method are purified towards energy-saving high sulfur-containing natural gas
CN106777468A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur content natural gas desulfurization process strong tracking evolutionary Modeling method
CN106777465A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur-containing natural gas purify technique dynamic evolutionary modeling and energy conservation optimizing method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
基于大数据的高含硫天然气脱硫工艺优化;辜小花等;《天然气工业》;20161231;第36卷(第9期);全文 *
基于神经网络的非线性系统自适应优化控制研究;罗艳红;《中国博士学位论文全文数据库信息科技辑》;20110615;第9-14，42-48页及图1.1,1.3-1.4,图3.1 *
平方根UKF神经网络及其在预测中的应用;黄冬民等;《昆明理工大学学报（理工版）》;20071231;第32卷(第3期);全文 *
自适应动态规划综述;张化光;《自动化学报》;20131231;第39卷(第4期);全文 *

Also Published As

Publication number Publication date
CN107908108A (en) 2018-04-13

Similar Documents

Publication Publication Date Title
CN109062053B (en) Denitration ammonia injection control method based on multivariate correction
CN106022521B (en) Short-term load prediction method of distributed BP neural network based on Hadoop architecture
CN108897286B (en) Fault detection method based on distributed nonlinear dynamic relation model
CN112085277B (en) SCR denitration system prediction model optimization method based on machine learning
CN106777866B (en) Energy-saving consumption-reducing-oriented modeling and optimizing method for high-sulfur-content natural gas purification process
CN107703760B (en) Natural gas absorption tower desulfurization process control method based on RBF and GDHP
CN105487376A (en) Optimal control method based on data-driven single network structure
CN109143872A (en) A kind of continuous stirred tank reactor course control method for use based on event triggering GDHP
CN103399488B (en) Multiple Model Control Method based on self study
CN109143853B (en) Self-adaptive control method for liquid level of fractionating tower in petroleum refining process
CN111158237A (en) Industrial furnace temperature multi-step prediction control method based on neural network
CN112742187A (en) Method and device for controlling pH value in desulfurization system
CN102393645A (en) Control method of high-speed electro-hydraulic proportional governing system
CN107908108B (en) Natural gas absorption tower desulfurization process control method based on UKF and GDHP
CN106777466B (en) Dynamic evolution modeling method of high-sulfur natural gas purification process based on ST-UPFNN algorithm
CN104616072A (en) Method for improving concentration of glutamic acid fermented product based on interval optimization
CN107831666B (en) Natural gas absorption tower desulfurization process control method based on RBF and ADDHP
CN107885084B (en) Natural gas absorption tower desulfurization process control method based on RBF and ADHDP
CN109240362A (en) A kind of method for adjusting pH value of waste water based on Fractional Order PID Controller
CN107885083B (en) Natural gas absorption tower desulfurization process control method based on UKF and ADHDP
CN107831665B (en) Natural gas absorption tower desulfurization process control method based on UKF and ADDHP
CN114924489B (en) Model autonomous learning method suitable for process industry prediction control
CN113964884A (en) Power grid active frequency regulation and control method based on deep reinforcement learning
CN113741182B (en) Sewage treatment process control method based on generalized value iteration
Guo et al. An improved generalized predictive control algorithm based on PID

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