[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN108509692B - A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm - Google Patents

A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm Download PDF

Info

Publication number
CN108509692B
CN108509692B CN201810201222.8A CN201810201222A CN108509692B CN 108509692 B CN108509692 B CN 108509692B CN 201810201222 A CN201810201222 A CN 201810201222A CN 108509692 B CN108509692 B CN 108509692B
Authority
CN
China
Prior art keywords
neural network
network model
state
state variable
covariance
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
CN201810201222.8A
Other languages
Chinese (zh)
Other versions
CN108509692A (en
Inventor
辜小花
张堃
唐海红
王甜
张兴
宋鸿飞
侯松
聂玲
李太福
邱奎
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 CN201810201222.8A priority Critical patent/CN108509692B/en
Publication of CN108509692A publication Critical patent/CN108509692A/en
Application granted granted Critical
Publication of CN108509692B publication Critical patent/CN108509692B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Geometry (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Treating Waste Gases (AREA)

Abstract

本发明公开了一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法，包括：S1：选择影响脱硫效率的工艺参数和脱硫单元的性能指标；S2：采集预设时间的所述工艺参数和所述性能指标的数据，剔除误差样本后形成样本集；S3：对样本集进行归一化形成归一化样本集，并从中选取训练样本和测试样本；S4：基于训练样本构建神经网络模型和初始状态变量；S5：利用MiUKFNN算法估计最优状态变量；S6：将最优状态变量作为神经网络模型的连接权值和阈值，即获得权值阈值更新后的神经网络模型；S7：得到预测结果，将预测结果与所述测试样本中的实际输出进行比较，如小于预设误差精度，神经网络模型有效；否则重复上述步骤至比较结果小于预设误差精度。

Figure 201810201222

The invention discloses a high-sulfur natural gas desulfurization process modeling method based on MiUKFNN algorithm, comprising: S1: selecting process parameters affecting desulfurization efficiency and performance indicators of desulfurization units; S2: collecting the process parameters and For the data of the performance indicators, a sample set is formed after removing error samples; S3: normalizing the sample set to form a normalized sample set, and selecting training samples and test samples from it; S4: constructing a neural network model and a test sample based on the training samples. Initial state variable; S5: Use MiUKFNN algorithm to estimate the optimal state variable; S6: Use the optimal state variable as the connection weight and threshold of the neural network model, that is, obtain the neural network model after updating the weight threshold; S7: Obtain the prediction result , compare the predicted result with the actual output in the test sample, if it is less than the preset error accuracy, the neural network model is valid; otherwise, repeat the above steps until the comparison result is less than the preset error accuracy.

Figure 201810201222

Description

一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

技术领域technical field

本发明涉及高含硫天然气净化技术领域，更为具体地，涉及一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法。The invention relates to the technical field of high-sulfur natural gas purification, and more particularly, to a high-sulfur natural gas desulfurization process modeling method based on MiUKFNN algorithm.

背景技术Background technique

高含硫天然气酸性组分含量比常规天然气高出数倍，其脱硫过程胺液循环量大、工艺流程复杂、能耗高。统计表明，脱硫单元能耗占高含硫天然气净化厂总能耗50％以上，其单位综合能耗高达1729.3MJ·t-1，属于高耗能单元。对大型净化厂而言，通过脱硫单元优化可降低能耗5％～10％。此外，高含硫天然气酸性组分浓度高，经过净化后的产品气量相对原料气流量有显著下降。为此，对高含硫天然气脱硫过程进行工艺优化，实现节能降耗，提高产率和气体加工经济效益是十分必要的。The content of acidic components in high-sulfur natural gas is several times higher than that of conventional natural gas, and its desulfurization process has a large circulating amount of amine liquid, complicated process flow and high energy consumption. Statistics show that the energy consumption of the desulfurization unit accounts for more than 50% of the total energy consumption of the high-sulfur natural gas purification plant, and the unit comprehensive energy consumption is as high as 1729.3MJ·t-1, which is a high energy consumption unit. For large-scale purification plants, the energy consumption can be reduced by 5% to 10% through the optimization of the desulfurization unit. In addition, the high concentration of acid components of high-sulfur natural gas, the purified product gas volume is significantly lower than the feed gas flow rate. For this reason, it is very necessary to optimize the process of high-sulfur natural gas desulfurization, realize energy saving and consumption reduction, and improve the yield and economic benefits of gas processing.

发明内容SUMMARY OF THE INVENTION

本发明的目的在于克服现有技术的不足，提供一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法。The purpose of the present invention is to overcome the deficiencies of the prior art and provide a high-sulfur natural gas desulfurization process modeling method based on the MiUKFNN algorithm.

本发明的目的是这样实现的：The object of the present invention is achieved in this way:

一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法，其特征在于，本方法包括以下步骤：A high-sulfur natural gas desulfurization process modeling method based on MiUKFNN algorithm, characterized in that the method comprises the following steps:

步骤S1：选择影响脱硫效率的工艺参数和脱硫单元的性能指标；Step S1: selecting the process parameters that affect the desulfurization efficiency and the performance index of the desulfurization unit;

步骤S2：数据采集及预处理，采集预设时间的所述影响脱硫效率的工艺参数和所述脱硫单元的性能指标的数据，剔除误差样本后形成样本集[X，Y]；Step S2: data collection and preprocessing, collecting the data of the process parameters affecting the desulfurization efficiency and the performance index of the desulfurization unit at a preset time, and forming a sample set [X, Y] after eliminating the error samples;

步骤S3：对样本集[X，Y]进行归一化，形成归一化样本集

Figure BDA0001594546670000021
取所述归一化样本集
Figure BDA0001594546670000022
中80％作为训练样本，剩余部分作为测试样本；Step S3: Normalize the sample set [X, Y] to form a normalized sample set
Figure BDA0001594546670000021
Take the normalized sample set
Figure BDA0001594546670000022
80% of them are used as training samples, and the rest are used as test samples;

步骤S4：基于所述训练样本构建神经网络模型和所述神经网络模型的初始状态变量θk，以及，将所述训练样本中的

Figure BDA0001594546670000023
作为所述神经网络模型的输入，将所述训练样本中的
Figure BDA0001594546670000024
作为所述神经网络模型的输出；Step S4: constructing a neural network model and an initial state variable θ k of the neural network model based on the training samples, and
Figure BDA0001594546670000023
As the input of the neural network model, the
Figure BDA0001594546670000024
as the output of the neural network model;

所述神经网络模型为：The neural network model is:

Figure BDA0001594546670000025
Figure BDA0001594546670000025

Figure BDA0001594546670000026
Figure BDA0001594546670000026

其中，

Figure BDA0001594546670000027
为所述训练样本的矢量样本值，并作为所述神经网络模型的输入；zj作为所述神经网络模型的隐含层输出；yd作为所述神经网络模型的输出层输出；wij为神经网络模型的输入层到隐含层的神经元的连接权值；
Figure BDA0001594546670000028
为神经网络模型的输入层到隐含层的神经元的阈值；vjd为所述神经网络模型的隐含层到输出层的神经元的连接权值，
Figure BDA0001594546670000029
为所述神经网络模型的隐含层到输出层的神经元的阈值，i＝1，2，…，m；m为神经网络模型的输入层的神经元的数量，s为神经网络模型的隐含层的神经元的数量，h为神经网络模型的输出层的神经元的数量；in,
Figure BDA0001594546670000027
is the vector sample value of the training sample and is used as the input of the neural network model; z j is the output of the hidden layer of the neural network model; y d is the output of the output layer of the neural network model; w ij is The connection weights of neurons from the input layer of the neural network model to the hidden layer;
Figure BDA0001594546670000028
is the threshold value of the neurons from the input layer of the neural network model to the hidden layer; vjd is the connection weight of the neurons from the hidden layer to the output layer of the neural network model,
Figure BDA0001594546670000029
is the threshold of neurons from the hidden layer to the output layer of the neural network model, i=1, 2, ..., m; m is the number of neurons in the input layer of the neural network model, and s is the hidden layer of the neural network model. The number of neurons in the containing layer, h is the number of neurons in the output layer of the neural network model;

应用于神经网络模型各层神经元的非线性激活函数为：The nonlinear activation function applied to the neurons in each layer of the neural network model is:

Figure BDA00015945466700000210
Figure BDA00015945466700000210

fo(x）＝x (4)f o (x) = x (4)

所述初始状态变量为：

Figure BDA00015945466700000211
The initial state variables are:
Figure BDA00015945466700000211

步骤S5：利用MiUKFNN算法估计所述神经网络模型的最优状态变量；Step S5: utilize MiUKFNN algorithm to estimate the optimal state variable of described neural network model;

步骤S6：将所述最优状态变量作为所述神经网络模型的wij、vjd、

Figure BDA00015945466700000212
和
Figure BDA00015945466700000213
对神经网络模型的公式进行更新，获得训练样本更新后的神经网络模型；Step S6: take the optimal state variables as w ij , v jd , and
Figure BDA00015945466700000212
and
Figure BDA00015945466700000213
Update the formula of the neural network model to obtain the updated neural network model of the training sample;

步骤S7：将所述测试样本中的

Figure BDA00015945466700000311
输入到更新后的神经网络模型，得到预测结果，将所述预测结果与所述测试样本中的实际输出
Figure BDA00015945466700000310
进行比较，如果比较结果小于预设误差值，所构建的神经网络模型有效；否则重复上述步骤S1-S7，直至所述比较结果小于所述预设误差值为止。Step S7: put the test samples in the
Figure BDA00015945466700000311
Input into the updated neural network model, get the prediction result, compare the prediction result with the actual output in the test sample
Figure BDA00015945466700000310
The comparison is performed, and if the comparison result is smaller than the preset error value, the constructed neural network model is valid; otherwise, the above steps S1-S7 are repeated until the comparison result is smaller than the preset error value.

优选地，所述步骤S5包括：Preferably, the step S5 includes:

步骤S51：在建立的所述神经网络模型中，将神经网络模型的权值和阈值组成的参数向量视为MiUKFNN算法所需的状态方程，将神经网络模型的输出视为MiUKFNN算法所需的量测方程：Step S51: In the established neural network model, the parameter vector composed of the weights and thresholds of the neural network model is regarded as the state equation required by the MiUKFNN algorithm, and the output of the neural network model is regarded as the quantity required by the MiUKFNN algorithm. Measuring equation:

θk＝θk-1+ηk (5)θ k = θ k-1 + η k (5)

Figure BDA0001594546670000031
Figure BDA0001594546670000031

其中，

Figure BDA0001594546670000032
为神经网络模型的输入，yk为神经网络模型的输出，
Figure BDA0001594546670000033
是参数化的非线性函数，ηk是过程噪声，μk是测量噪声；in,
Figure BDA0001594546670000032
is the input of the neural network model, y k is the output of the neural network model,
Figure BDA0001594546670000033
is the parameterized nonlinear function, η k is the process noise, μ k is the measurement noise;

对状态方程和量测方程进行初始化，计算状态变量估计以及其协方差：Initialize the state and measurement equations, and compute the state variable estimates and their covariances:

Figure BDA0001594546670000034
Figure BDA0001594546670000034

Figure BDA0001594546670000035
Figure BDA0001594546670000035

其中：

Figure BDA0001594546670000036
in:
Figure BDA0001594546670000036

步骤S52：运用减少Sigma点集方法对所述初始状态变量θk进行Sigma采样，获得n+1个采样点以及权重系数，随机变量

Figure BDA0001594546670000037
具有均值
Figure BDA0001594546670000038
和协方差矩阵PXX＞0，则：Step S52: Sigma sampling is performed on the initial state variable θ k by using the method of reducing the Sigma point set to obtain n+1 sampling points and weight coefficients, random variables
Figure BDA0001594546670000037
has mean
Figure BDA0001594546670000038
and covariance matrix P XX > 0, then:

Figure BDA0001594546670000039
Figure BDA0001594546670000039

Wweight＝[W ωn+1] (10)W weight = [W ω n+1 ] (10)

其中：in:

Figure BDA0001594546670000041
Figure BDA0001594546670000041

Figure BDA0001594546670000042
Figure BDA0001594546670000042

步骤S53：状态更新，通过离散时间非线性系统的状态方程将每个采样点的k时刻的最优状态变量的状态估计变换为k+1时刻的状态变量的状态估计

Figure BDA0001594546670000043
并通过合并k+1时刻的状态估计
Figure BDA0001594546670000044
的向量，获得k+1时刻的状态变量的状态先验估计
Figure BDA0001594546670000045
和协方差
Figure BDA0001594546670000046
其中，所述状态估计
Figure BDA0001594546670000047
为：Step S53: state update, transform the state estimate of the optimal state variable at time k of each sampling point into the state estimate of the state variable at time k+1 through the state equation of the discrete-time nonlinear system
Figure BDA0001594546670000043
And by merging the state estimation at time k+1
Figure BDA0001594546670000044
The vector of , obtain the state a priori estimate of the state variable at time k+1
Figure BDA0001594546670000045
and covariance
Figure BDA0001594546670000046
where the state estimate
Figure BDA0001594546670000047
for:

Figure BDA0001594546670000048
Figure BDA0001594546670000048

其中，ηk为过程噪声，其协方差矩阵Qk为cov(wk，wj)＝Qkδkj，

Figure BDA0001594546670000049
Among them, η k is the process noise, and its covariance matrix Q k is cov(w k , w j )=Q k δ kj ,
Figure BDA0001594546670000049

所述状态先验估计

Figure BDA00015945466700000410
为：The state prior estimates
Figure BDA00015945466700000410
for:

Figure BDA00015945466700000411
Figure BDA00015945466700000411

所述状态变量的协方差

Figure BDA00015945466700000412
为：the covariance of the state variables
Figure BDA00015945466700000412
for:

Figure BDA00015945466700000413
Figure BDA00015945466700000413

步骤S54：量测更新，通过离散时间非线性系统的量测方程建立k时刻的状态变量的状态估计

Figure BDA00015945466700000414
和k时刻的量测预测估计
Figure BDA00015945466700000415
之间的联系以完成量测预测，并估计k时刻的量测预测的协方差
Figure BDA00015945466700000416
以及k时刻的状态变量和量测预测之间的协方差
Figure BDA00015945466700000417
Step S54: measurement update, establishing the state estimation of the state variable at time k through the measurement equation of the discrete-time nonlinear system
Figure BDA00015945466700000414
and the measurement prediction estimate at time k
Figure BDA00015945466700000415
to complete the measurement forecast and estimate the covariance of the measurement forecast at time k
Figure BDA00015945466700000416
and the covariance between the state variable and the measurement prediction at time k
Figure BDA00015945466700000417

所述k时刻的量测预测的均值

Figure BDA00015945466700000418
为：The mean value of the measurement prediction at the k time
Figure BDA00015945466700000418
for:

Figure BDA00015945466700000419
Figure BDA00015945466700000419

其中，

Figure BDA00015945466700000420
为神经网络模型预测输出，由神经网络模型的公式得出；in,
Figure BDA00015945466700000420
Predict the output for the neural network model, which is obtained by the formula of the neural network model;

所述k时刻的量测预测的协方差

Figure BDA0001594546670000051
为：The covariance of the measurement prediction at the k time instant
Figure BDA0001594546670000051
for:

Figure BDA0001594546670000052
Figure BDA0001594546670000052

所述k时刻的状态变量和量测预测之间的协方差

Figure BDA0001594546670000053
为：The covariance between the state variable at time k and the measurement prediction
Figure BDA0001594546670000053
for:

Figure BDA0001594546670000054
Figure BDA0001594546670000054

步骤S55：通过建立协方差

Figure BDA0001594546670000055
和协方差
Figure BDA0001594546670000056
的关系，更新k时刻的状态变量的状态估计和协方差；Step S55: By establishing covariance
Figure BDA0001594546670000055
and covariance
Figure BDA0001594546670000056
, update the state estimate and covariance of the state variable at time k;

所述协方差之间的关系是：The relationship between the covariances is:

Figure BDA0001594546670000057
Figure BDA0001594546670000057

通过上述关系对k+1时刻的状态变量的状态估计和协方差进行修正：The state estimates and covariances of the state variables at time k+1 are corrected by the above relationship:

Figure BDA0001594546670000058
Figure BDA0001594546670000058

Figure BDA0001594546670000059
Figure BDA0001594546670000059

步骤S56：将获得的修正后k+1时刻的状态变量

Figure BDA00015945466700000510
重组神经网络模型，并计算此时神经网络模型的预测输出与实际输出之间的误差，如果小于既设精度要求，则输出所述神经网络模型的最优状态变量
Figure BDA00015945466700000511
反之，重新进入步骤S51。Step S56: the obtained corrected state variable at time k+1
Figure BDA00015945466700000510
Reorganize the neural network model, and calculate the error between the predicted output of the neural network model and the actual output at this time, if it is less than the preset accuracy requirement, output the optimal state variable of the neural network model
Figure BDA00015945466700000511
On the contrary, re-enter step S51.

优选地，所述影响脱硫效率的工艺参数包括进入尾气吸收塔的贫胺液流量、进入二级吸收塔的贫胺液流量、原料气处理量、尾气单元返回脱硫单元的半富胺液流量、一级吸收塔胺液入塔温度、二级吸收塔胺液入塔温度、闪蒸罐压力、蒸汽预热器的蒸汽消耗量、一个重沸器的蒸汽消耗量、另一个重沸器的蒸汽消耗量；所述脱硫单元的性能指标包括净化气中H2S和CO2的浓度。Preferably, the process parameters affecting desulfurization efficiency include the flow rate of lean amine liquid entering the tail gas absorption tower, the flow rate of lean amine liquid entering the secondary absorption tower, the processing capacity of raw material gas, the flow rate of semi-rich amine liquid returning from the tail gas unit to the desulfurization unit, Amine liquid inlet temperature of primary absorption tower, amine liquid inlet temperature of secondary absorption tower, flash tank pressure, steam consumption of steam preheater, steam consumption of one reboiler, steam of another reboiler consumption; the performance indicators of the desulfurization unit include the concentration of H 2 S and CO 2 in the purified gas.

优选地，步骤S3中，随机选取所述归一化样本集

Figure BDA00015945466700000512
中80％的样本作为训练样本，而剩余的20％样本作为测试样本。Preferably, in step S3, the normalized sample set is randomly selected
Figure BDA00015945466700000512
80% of the samples are used as training samples, and the remaining 20% of samples are used as test samples.

由于采用了上述技术方案，本发明相对于现有技术能够节能降耗，提高产率和气体加工经济效益。Due to the adoption of the above technical solution, the present invention can save energy and reduce consumption compared with the prior art, and improve the yield and the economic benefit of gas processing.

附图说明Description of drawings

图1a、图1b为训练样本的拟合精度图；Figure 1a and Figure 1b are the fitting accuracy diagrams of the training samples;

图2a、图2b为训练样本的均方误差MSE以及最大绝对误差MAE；Figure 2a and Figure 2b are the mean square error MSE and the maximum absolute error MAE of the training samples;

图3a、图3b为测试样本的测试精度图；Fig. 3a, Fig. 3b are the test accuracy figures of the test sample;

图4a、图4b为测试样本的误差统计图；Fig. 4a, Fig. 4b are the error statistics of the test sample;

图5a、图5b为MiUKFNN模型关于H2S和CO2各自浓度的可靠性的Williams图。Figures 5a and 5b are Williams plots of the reliability of the MiUKFNN model with respect to the respective concentrations of H 2 S and CO 2 .

具体实施方式Detailed ways

MiUKFNN解释：Minimum unscented Kalman filter neural network，减少Sigma点的无迹卡尔曼滤波神经网络。MiUKFNN explained: Minimum unscented Kalman filter neural network, an unscented Kalman filter neural network that reduces Sigma points.

一种基于MiUKFNN算法的高含硫天然气脱硫工艺建模方法，本方法包括以下步骤：A high-sulfur natural gas desulfurization process modeling method based on MiUKFNN algorithm, the method comprises the following steps:

步骤S1：选择影响脱硫效率的工艺参数和脱硫单元的性能指标；其中，所述影响脱硫效率的工艺参数包括进入尾气吸收塔的贫胺液流量、进入二级吸收塔的贫胺液流量、原料气处理量、尾气单元返回脱硫单元的半富胺液流量、一级吸收塔胺液入塔温度、二级吸收塔胺液入塔温度、闪蒸罐压力、一个重沸器的蒸汽消耗量、另一个重沸器的蒸汽消耗量和蒸汽预热器的蒸汽消耗量；脱硫单元的性能指标包括净化气中H2S和CO2的浓度，如表1所示：Step S1: Select the process parameters that affect the desulfurization efficiency and the performance index of the desulfurization unit; wherein, the process parameters that affect the desulfurization efficiency include the flow rate of the lean amine liquid entering the tail gas absorption tower, the flow rate of the lean amine liquid entering the secondary absorption tower, and the raw materials. gas treatment capacity, the flow rate of semi-rich amine liquid returned from the tail gas unit to the desulfurization unit, the inlet temperature of the amine liquid in the primary absorption tower, the inlet temperature of the amine liquid in the secondary absorption tower, the pressure of the flash tank, the steam consumption of a reboiler, The steam consumption of the other reboiler and the steam consumption of the steam preheater; the performance indicators of the desulfurization unit include the concentration of H 2 S and CO 2 in the purified gas, as shown in Table 1:

表1实验参数列表Table 1 List of experimental parameters

Figure BDA0001594546670000071
Figure BDA0001594546670000071

步骤S2：：数据采集及预处理：采集预设时间的所述工艺参数和所述性能指标的数据，剔除误差样本后形成样本集[X，Y]，如表2所示：Step S2: data collection and preprocessing: collect the data of the process parameters and the performance indicators at the preset time, and form a sample set [X, Y] after eliminating the error samples, as shown in Table 2:

表2原始样本集Table 2 Original sample set

Figure BDA0001594546670000072
Figure BDA0001594546670000072

步骤S3：对样本集[X，Y]进行归一化，形成归一化样本集

Figure BDA0001594546670000073
取所述归一化样本集
Figure BDA0001594546670000074
中前80％的样本作为训练样本，而剩余的20％样本作为测试样本。Step S3: Normalize the sample set [X, Y] to form a normalized sample set
Figure BDA0001594546670000073
Take the normalized sample set
Figure BDA0001594546670000074
The top 80% of the samples are used as training samples, and the remaining 20% of samples are used as test samples.

步骤S4：基于所述训练样本构建神经网络模型和所述神经网络模型的初始状态变量θk，以及，将所述训练样本中的

Figure BDA0001594546670000075
作为所述神经网络模型的输入，将所述训练样本中的
Figure BDA0001594546670000076
作为所述神经网络模型的输出；Step S4: constructing a neural network model and an initial state variable θ k of the neural network model based on the training samples, and
Figure BDA0001594546670000075
As the input of the neural network model, the
Figure BDA0001594546670000076
as the output of the neural network model;

所述神经网络模型为：The neural network model is:

Figure BDA0001594546670000081
Figure BDA0001594546670000081

Figure BDA0001594546670000082
Figure BDA0001594546670000082

其中，

Figure BDA0001594546670000083
为所述训练样本的矢量样本值，并作为所述神经网络模型的输入；zj作为所述神经网络模型的隐含层输出；yd作为所述神经网络模型的输出层输出；wij为神经网络模型的输入层到隐含层的神经元的连接权值；
Figure BDA0001594546670000084
为神经网络模型的输入层到隐含层的神经元的阈值；vjd为所述神经网络模型的隐含层到输出层的神经元的连接权值，
Figure BDA0001594546670000085
为所述神经网络模型的隐含层到输出层的神经元的阈值，i＝1，2，…，m；m为神经网络模型的输入层的神经元的数量，s为神经网络模型的隐含层的神经元的数量，h为神经网络模型的输出层的神经元的数量；in,
Figure BDA0001594546670000083
is the vector sample value of the training sample and is used as the input of the neural network model; z j is the output of the hidden layer of the neural network model; y d is the output of the output layer of the neural network model; w ij is The connection weights of neurons from the input layer of the neural network model to the hidden layer;
Figure BDA0001594546670000084
is the threshold value of the neurons from the input layer of the neural network model to the hidden layer; vjd is the connection weight of the neurons from the hidden layer to the output layer of the neural network model,
Figure BDA0001594546670000085
is the threshold of neurons from the hidden layer to the output layer of the neural network model, i=1, 2, ..., m; m is the number of neurons in the input layer of the neural network model, and s is the hidden layer of the neural network model. The number of neurons in the containing layer, h is the number of neurons in the output layer of the neural network model;

应用于神经网络模型各层神经元的非线性激活函数为：The nonlinear activation function applied to the neurons in each layer of the neural network model is:

Figure BDA0001594546670000086
Figure BDA0001594546670000086

fo(x)＝x (4)f o (x)=x (4)

所述初始状态变量为：

Figure BDA0001594546670000087
The initial state variables are:
Figure BDA0001594546670000087

步骤S5：利用MiUKFNN算法估计所述神经网络模型的最优状态变量；Step S5: utilize MiUKFNN algorithm to estimate the optimal state variable of described neural network model;

本发明利用MiUKFNN算法估计神经网络模型的状态变量，以达到连接权值、阈值的不断调整，直到满足要求。将得到的最优状态变量的状态估计作为上述所建立神经网络模型的连接权值、阈值。需要说明的是，该连接权值、阈值为通过MiUKFNN算法调整后的连接权值、阈值，也是上述所建立的神经网络模型的全部连接权值与阈值，包括

Figure BDA0001594546670000088
The present invention uses the MiUKFNN algorithm to estimate the state variables of the neural network model, so as to achieve continuous adjustment of connection weights and thresholds until the requirements are met. The obtained state estimation of the optimal state variable is used as the connection weight and threshold of the neural network model established above. It should be noted that the connection weights and thresholds are the connection weights and thresholds adjusted by the MiUKFNN algorithm, and are also all connection weights and thresholds of the neural network model established above, including
Figure BDA0001594546670000088

利用MiUKFNN算法估计神经网络模型的最优状态变量的过程包括：The process of using the MiUKFNN algorithm to estimate the optimal state variables of the neural network model includes:

步骤S51：在建立的所述神经网络模型中，将神经网络模型的权值和阈值组成的参数向量视为MiUKFNN算法所需的状态方程，将神经网络模型的输出视为MiUKFNN算法所需的量测方程：Step S51: In the established neural network model, the parameter vector composed of the weights and thresholds of the neural network model is regarded as the state equation required by the MiUKFNN algorithm, and the output of the neural network model is regarded as the quantity required by the MiUKFNN algorithm. Measuring equation:

θk＝θk-1+ηk (5)θ k = θ k-1 + η k (5)

Figure BDA0001594546670000091
Figure BDA0001594546670000091

其中，

Figure BDA0001594546670000092
为神经网络模型的输入，yk为神经网络模型的输出，
Figure BDA0001594546670000093
是参数化的非线性函数，ηk是过程噪声，μk是测量噪声；in,
Figure BDA0001594546670000092
is the input of the neural network model, y k is the output of the neural network model,
Figure BDA0001594546670000093
is the parameterized nonlinear function, η k is the process noise, μ k is the measurement noise;

对状态方程和量测方程进行初始化，计算状态变量估计以及其协方差：Initialize the state and measurement equations, and compute the state variable estimates and their covariances:

Figure BDA0001594546670000094
Figure BDA0001594546670000094

Figure BDA0001594546670000095
Figure BDA0001594546670000095

其中：

Figure BDA0001594546670000096
in:
Figure BDA0001594546670000096

步骤S52：运用减少Sigma点集方法对所述初始状态变量θk进行Sigma采样，获得n+1个采样点以及权重系数，随机变量

Figure BDA0001594546670000097
具有均值
Figure BDA0001594546670000098
和协方差矩阵PXX＞0，则：Step S52: Sigma sampling is performed on the initial state variable θ k by using the method of reducing the Sigma point set to obtain n+1 sampling points and weight coefficients, random variables
Figure BDA0001594546670000097
has mean
Figure BDA0001594546670000098
and covariance matrix P XX > 0, then:

Figure BDA0001594546670000099
Figure BDA0001594546670000099

Wweight＝[W ωn+1] (10)W weight = [W ω n+1 ] (10)

其中：in:

Figure BDA00015945466700000910
Figure BDA00015945466700000910

Figure BDA00015945466700000911
Figure BDA00015945466700000911

步骤S53：状态更新，通过离散时间非线性系统的状态方程将每个采样点的k时刻的最优状态变量的状态估计变换为k+1时刻的状态变量的状态估计

Figure BDA00015945466700000912
并通过合并k+1时刻的状态估计
Figure BDA00015945466700000913
的向量，获得k+1时刻的状态变量的状态先验估计
Figure BDA0001594546670000101
和协方差
Figure BDA0001594546670000102
其中，所述状态估计
Figure BDA0001594546670000103
为：Step S53: state update, transform the state estimate of the optimal state variable at time k of each sampling point into the state estimate of the state variable at time k+1 through the state equation of the discrete-time nonlinear system
Figure BDA00015945466700000912
And by merging the state estimation at time k+1
Figure BDA00015945466700000913
The vector of , obtain the state a priori estimate of the state variable at time k+1
Figure BDA0001594546670000101
and covariance
Figure BDA0001594546670000102
where the state estimate
Figure BDA0001594546670000103
for:

Figure BDA0001594546670000104
Figure BDA0001594546670000104

其中，ηk为过程噪声，其协方差矩阵Qk为cov(wk，wj)＝Qkδkj，

Figure BDA0001594546670000105
Among them, η k is the process noise, and its covariance matrix Q k is cov(w k , w j )=Q k δ kj ,
Figure BDA0001594546670000105

所述状态先验估计

Figure BDA0001594546670000106
为：The state prior estimates
Figure BDA0001594546670000106
for:

Figure BDA0001594546670000107
Figure BDA0001594546670000107

所述状态变量的协方差

Figure BDA0001594546670000108
为：the covariance of the state variables
Figure BDA0001594546670000108
for:

Figure BDA0001594546670000109
Figure BDA0001594546670000109

步骤S54：量测更新，通过离散时间非线性系统的量测方程建立k时刻的状态变量的状态估计

Figure BDA00015945466700001010
和k时刻的量测预测估计
Figure BDA00015945466700001011
之间的联系以完成量测预测，并估计k时刻的量测预测的协方差
Figure BDA00015945466700001012
以及k时刻的状态变量和量测预测之间的协方差
Figure BDA00015945466700001013
Step S54: measurement update, establishing the state estimation of the state variable at time k through the measurement equation of the discrete-time nonlinear system
Figure BDA00015945466700001010
and the measurement prediction estimate at time k
Figure BDA00015945466700001011
to complete the measurement forecast and estimate the covariance of the measurement forecast at time k
Figure BDA00015945466700001012
and the covariance between the state variable and the measurement prediction at time k
Figure BDA00015945466700001013

所述k时刻的量测预测的均值

Figure BDA00015945466700001014
为：The mean value of the measurement prediction at the k time
Figure BDA00015945466700001014
for:

Figure BDA00015945466700001015
Figure BDA00015945466700001015

其中，

Figure BDA00015945466700001016
为神经网络模型预测输出，由公式(1)与公式(2)得出；in,
Figure BDA00015945466700001016
is the predicted output of the neural network model, which is obtained by formula (1) and formula (2);

所述k时刻的量测预测的协方差

Figure BDA00015945466700001017
为：The covariance of the measurement prediction at the k time instant
Figure BDA00015945466700001017
for:

Figure BDA00015945466700001018
Figure BDA00015945466700001018

所述k时刻的状态变量和量测预测之间的协方差

Figure BDA00015945466700001019
为：The covariance between the state variable at time k and the measurement prediction
Figure BDA00015945466700001019
for:

Figure BDA00015945466700001020
Figure BDA00015945466700001020

步骤S55：通过建立协方差

Figure BDA00015945466700001021
和协方差
Figure BDA00015945466700001022
的关系，更新k时刻的状态变量的状态估计和协方差；Step S55: By establishing covariance
Figure BDA00015945466700001021
and covariance
Figure BDA00015945466700001022
, update the state estimate and covariance of the state variable at time k;

所述协方差之间的关系是：The relationship between the covariances is:

Figure BDA0001594546670000111
Figure BDA0001594546670000111

通过上述关系对k+1时刻的状态变量的状态估计和协方差进行修正：The state estimates and covariances of the state variables at time k+1 are corrected by the above relationship:

Figure BDA0001594546670000112
Figure BDA0001594546670000112

Figure BDA0001594546670000113
Figure BDA0001594546670000113

步骤S56：将获得的修正后k+1时刻的状态变量

Figure BDA0001594546670000114
重组BP神经网络模型，并计算此时神经网络模型的预测输出与实际输出之间的误差，如果小于既设精度要求，则输出所述神经网络模型的最优状态变量
Figure BDA0001594546670000115
反之，重新进入步骤S51。Step S56: the obtained corrected state variable at time k+1
Figure BDA0001594546670000114
Reorganize the BP neural network model, and calculate the error between the predicted output and the actual output of the neural network model at this time. If it is less than the preset accuracy requirement, output the optimal state variable of the neural network model.
Figure BDA0001594546670000115
On the contrary, re-enter step S51.

步骤S6：将所述最优状态变量作为所述神经网络模型的wij、vjd、

Figure BDA0001594546670000116
和
Figure BDA0001594546670000117
对公式(1)和公式(2)进行更新，获得训练样本更新后的神经网络模型；Step S6: take the optimal state variables as w ij , v jd , and
Figure BDA0001594546670000116
and
Figure BDA0001594546670000117
Update formula (1) and formula (2) to obtain the updated neural network model of the training sample;

步骤S7：将所述测试样本中的

Figure BDA0001594546670000118
输入到更新后的神经网络模型，得到预测结果，将所述预测结果与所述测试样本中的实际输出
Figure BDA0001594546670000119
进行比较，如果比较结果小于预设误差值，所构建的神经网络模型有效；否则重复上述步骤S1-S7，直至所述比较结果小于所述预设误差值为止。Step S7: put the test samples in the
Figure BDA0001594546670000118
Input into the updated neural network model, get the prediction result, compare the prediction result with the actual output in the test sample
Figure BDA0001594546670000119
The comparison is performed, and if the comparison result is smaller than the preset error value, the constructed neural network model is valid; otherwise, the above steps S1-S7 are repeated until the comparison result is smaller than the preset error value.

本发明通过几组测试得到如下的技术效果：The present invention obtains following technical effect through several groups of tests:

图1a-图1b比较了训练样本在BPNN、UKFNN、以及MiUKFNN的拟合精度图，其中，图1a影响脱硫效率的工艺参数为净化气H2S浓度，图1b影响脱硫效率的工艺参数为净化气CO2浓度。其中，图中所画45°的趋势线表示预测值等于实际值时产生的效果，图中的点越接近45°线，表明模型的预测值与实际值之间的误差越小。Figures 1a-1b compare the fitting accuracy graphs of the training samples in BPNN, UKFNN, and MiUKFNN. The process parameter affecting the desulfurization efficiency in Figure 1a is the concentration of purified gas H 2 S, and the process parameter affecting the desulfurization efficiency in Figure 1b is the purification process. Gas CO2 concentration. Among them, the 45° trend line drawn in the figure represents the effect when the predicted value is equal to the actual value. The closer the point in the figure is to the 45° line, the smaller the error between the predicted value of the model and the actual value.

图2a-图2b为以上三种方法下训练样本的均方误差MSE以及最大绝对误差MAE，其中图2a影响脱硫效率的工艺参数为净化气H2S浓度，图2b影响脱硫效率的工艺参数为净化气CO2浓度。Fig. 2a-Fig. 2b are the mean square error MSE and the maximum absolute error MAE of the training samples under the above three methods. The process parameter affecting the desulfurization efficiency in Fig. 2a is the H2S concentration of the purified gas, and the process parameter affecting the desulfurization efficiency in Fig. 2b is the purified gas. CO2 concentration.

图3a-图3b为测试样本的测试精度图，其中，图3a影响脱硫效率的工艺参数为净化气H2S浓度，图3b影响脱硫效率的工艺参数为净化气CO2浓度。Figures 3a-3b are test accuracy diagrams of the test samples, wherein the process parameter affecting the desulfurization efficiency in Figure 3a is the concentration of purified gas H 2 S, and the process parameter affecting the desulfurization efficiency in Figure 3b is the concentration of purified gas CO 2 .

图4a-图4b为测试样本的误差统计图，其中，图4a影响脱硫效率的工艺参数为净化气H2S浓度，图4b影响脱硫效率的工艺参数为净化气CO2浓度。Figures 4a-4b are error statistics of the test samples, wherein the process parameter affecting the desulfurization efficiency in Figure 4a is the concentration of purified gas H 2 S, and the process parameter affecting the desulfurization efficiency in Figure 4b is the concentration of purified gas CO 2 .

结果表明，三种方法输出的训练误差都很小，预测值与实际数据点之间的最小均方误差分别为0、0.0046和0、0.005。虽然这三种方法的CO2测试误差略高于训练误差，但它们一般处于很低的水平。在图4a中，H2S浓度的MSE值和MAE值分别为0和0.0024，图4b中CO2浓度的MSE值和MAE值分别为0和0.006。在所有情况下，MiUKFNN产生的误差最少。这表明，所提出的MiUKFNN模型是可行的，故所建模有效。The results show that the training errors output by the three methods are all small, and the minimum mean square errors between the predicted values and the actual data points are 0, 0.0046 and 0, 0.005, respectively. Although the CO2 test errors of these three methods are slightly higher than the training errors, they are generally at very low levels. In Fig. 4a, the MSE and MAE values of the H2S concentration are 0 and 0.0024, respectively, and the MSE and MAE values of the CO2 concentration in Fig. 4b are 0 and 0.006, respectively. In all cases, MiUKFNN produces the least error. This shows that the proposed MiUKFNN model is feasible, so the modeling is effective.

最后，为了检验MiUKFNN模型的可靠性，采用了由模型结果的残差、Williams图和HAT矩阵组成的杠杆方法来检测可能的离群点。图5a-图5b描述了MiUKFNN模型关于H2S和CO2各自浓度的可靠性的Williams图。图中可以看出没有出现超出范围的离群点，证明模型具有一定的可靠性。Finally, to test the reliability of the MiUKFNN model, a leverage method consisting of residuals of model results, Williams plots, and HAT matrices is employed to detect possible outliers. Figures 5a-5b depict Williams plots of the reliability of the MiUKFNN model with respect to the respective concentrations of H2S and CO2 . It can be seen from the figure that there are no outliers out of range, which proves that the model has certain reliability.

以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应所述以权利要求的保护范围为准。The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention. should be included within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims (3)

1. A high-sulfur natural gas desulfurization process modeling method based on a MiUKFNN algorithm is characterized by comprising the following steps:
step S1: selecting technological parameters influencing the desulfurization efficiency and performance indexes of a desulfurization unit;
step S2: acquiring and preprocessing data, acquiring the technological parameters affecting the desulfurization efficiency and the data of the performance indexes of the desulfurization unit within preset time, and removing error samples to form a sample set [ X, Y ];
step S3: for sample set [ X, Y]Carrying out normalization to form a normalized sample set
Figure FDA0003355704760000011
Taking the normalized sample set
Figure FDA0003355704760000012
The middle 80 percent is used as a training sample, and the rest part is used as a test sample;
step S4: constructing a neural network model based on the training samples and an initial state variable theta of the neural network modelkAnd, in the training sample
Figure FDA0003355704760000013
As input to the neural network model, the training samples
Figure FDA0003355704760000014
As an output of the neural network model;
the neural network model is as follows:
Figure FDA0003355704760000015
Figure FDA0003355704760000016
wherein,
Figure FDA0003355704760000017
vector sample values of the training samples are used as input of the neural network model; z is a radical ofjAs a hidden layer output of the neural network model; y isdAs an output layer output of the neural network model; w is aijConnecting weights of neurons from an input layer to a hidden layer of the neural network model;
Figure FDA0003355704760000018
a threshold value of a neuron from an input layer to a hidden layer of the neural network model; v. ofjdThe connection weights of the neurons of the hidden layer to the output layer of the neural network model,
Figure FDA0003355704760000019
a threshold of neurons from hidden layer to output layer of the neural network model, i ═ 1,2, …, m; m is the number of neurons of an input layer of the neural network model, s is the number of neurons of a hidden layer of the neural network model, and h is the number of neurons of an output layer of the neural network model;
the nonlinear activation function applied to each layer of neurons of the neural network model is as follows:
Figure FDA0003355704760000021
fo(x)＝x (4)
the initial state variables are:
Figure FDA0003355704760000022
step S5: estimating the optimal state variable of the neural network model by using a MiUKFNN algorithm;
the step S5 includes:
step S51: in the established neural network model, a parameter vector consisting of the weight and the threshold of the neural network model is regarded as a state equation required by the MiUKFNN algorithm, and the output of the neural network model is regarded as a measurement equation required by the MiUKFNN algorithm:
θk＝θk-1+ηk (5)
Figure FDA0003355704760000023
wherein,
Figure FDA0003355704760000024
as input to the neural network model, ykIs the output of the neural network model and,
Figure FDA0003355704760000025
is a parameterized non-linear function, ηkIs process noise, μkIs the measurement noise;
initializing a state equation and a measurement equation, and calculating state variable estimation and covariance thereof:
Figure FDA0003355704760000026
Figure FDA0003355704760000027
wherein:
Figure FDA0003355704760000028
step S52: applying the method of reducing Sigma point set to the initial state variable thetakCarrying out Sigma sampling to obtain n +1 sampling points, weight coefficients and random variables
Figure FDA0003355704760000029
Having a mean value
Figure FDA00033557047600000210
Sum covariance matrix PXX>0, then:
Figure FDA00033557047600000211
Wweight＝[W ωn+1] (10)
wherein:
Figure FDA0003355704760000031
Figure FDA0003355704760000032
step S53: updating state, namely converting the state estimation of the optimal state variable at the k moment of each sampling point into the state estimation of the state variable at the k +1 moment by using the state equation of the discrete time nonlinear system
Figure FDA0003355704760000033
And by combining the state estimates at the time k +1
Figure FDA0003355704760000034
To obtain a state prior estimate of the state variable at time k +1
Figure FDA0003355704760000035
Sum covariance
Figure FDA0003355704760000036
Wherein the state estimation
Figure FDA0003355704760000037
Comprises the following steps:
Figure FDA0003355704760000038
wherein eta iskAs process noise, its covariance matrix QkIs cov (w)k,wj)＝Qkδkj，
Figure FDA0003355704760000039
The state prior estimate
Figure FDA00033557047600000310
Comprises the following steps:
Figure FDA00033557047600000311
covariance of the state variable
Figure FDA00033557047600000312
Comprises the following steps:
Figure FDA00033557047600000313
step S54: measurement update, establishing state estimation of state variable at time k by using measurement equation of discrete time nonlinear system
Figure FDA00033557047600000314
And measured prediction estimation of time k
Figure FDA00033557047600000315
To complete the metrology prediction and estimate the covariance of the metrology prediction at time k
Figure FDA00033557047600000316
And covariance between state variables and metrology predictions at time k
Figure FDA00033557047600000317
Mean of the measured predictions of the k time
Figure FDA00033557047600000318
Comprises the following steps:
Figure FDA00033557047600000319
wherein,
Figure FDA00033557047600000320
predicting output of the neural network model, which is obtained by a formula of the neural network model;
covariance of metrology prediction at the time k
Figure FDA0003355704760000041
Comprises the following steps:
Figure FDA0003355704760000042
covariance between state variables and metrology predictions for the time k
Figure FDA0003355704760000043
Comprises the following steps:
Figure FDA0003355704760000044
step S55: by establishing covariance
Figure FDA0003355704760000045
Sum covariance
Figure FDA0003355704760000046
Updating the state estimation and covariance of the state variable at the moment k;
the relationship between the covariances is:
Figure FDA0003355704760000047
the state estimate and covariance of the state variables at time k +1 are corrected by the above relationship:
Figure FDA0003355704760000048
Figure FDA0003355704760000049
step S56: the obtained state variable at the time of k +1 after correction
Figure FDA00033557047600000410
Recombining the neural network model, calculating the error between the predicted output and the actual output of the neural network model at the moment, and outputting the optimal state variable of the neural network model if the error is less than the preset precision requirement
Figure FDA00033557047600000411
Otherwise, re-enter step S51;
step S6: using the optimal state variable as w of the neural network modelij、vjd、
Figure FDA00033557047600000412
And
Figure FDA00033557047600000413
updating the formula of the neural network model to obtain the neural network model after the training sample is updated;
step S7: in the test sample
Figure FDA00033557047600000414
Inputting the result into the updated neural network model to obtain a prediction result, and outputting the prediction result and the actual output in the test sample
Figure FDA00033557047600000415
Comparing, and if the comparison result is smaller than a preset error value, the constructed neural network model is effective; otherwise, repeating the above steps S1-S7 until the comparison result is less than the preset error value.
2. The MiUKFNN algorithm-based modeling method for desulfurization process of natural gas with high sulfur content as claimed in claim 1, wherein the process parameters affecting desulfurization efficiency include lean amine liquid flow entering the tail gas absorption tower, lean amine liquid flow entering the secondary absorption tower, raw gas treatment capacity, and half of tail gas unit return desulfurization unitRich amine liquid flow, the tower entering temperature of the amine liquid of the primary absorption tower, the tower entering temperature of the amine liquid of the secondary absorption tower, flash tank pressure, steam consumption of a steam preheater, steam consumption of one reboiler and steam consumption of the other reboiler; the performance index of the desulfurization unit includes H in the purge gas2S and CO2The concentration of (c).
3. The MiUKFNN algorithm-based modeling method for high-sulfur natural gas desulfurization process of claim 1, wherein in step S3, the normalized sample set is randomly selected
Figure FDA0003355704760000051
The middle 80% of the samples were used as training samples, and the remaining 20% were used as test samples.
CN201810201222.8A 2018-03-12 2018-03-12 A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm Active CN108509692B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810201222.8A CN108509692B (en) 2018-03-12 2018-03-12 A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810201222.8A CN108509692B (en) 2018-03-12 2018-03-12 A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Publications (2)

Publication Number Publication Date
CN108509692A CN108509692A (en) 2018-09-07
CN108509692B true CN108509692B (en) 2022-03-25

Family

ID=63377401

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810201222.8A Active CN108509692B (en) 2018-03-12 2018-03-12 A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Country Status (1)

Country Link
CN (1) CN108509692B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112216351B (en) * 2020-09-23 2024-02-13 成都佳华物链云科技有限公司 Desulfurization optimization method and device, electronic equipment and storage medium
CN113094992A (en) * 2021-04-12 2021-07-09 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 Desulfurizing tower optimization method based on neural network discrimination
CN113780579B (en) * 2021-09-09 2023-04-07 北京百度网讯科技有限公司 Natural gas purification foaming early warning model training method, device and equipment

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106777466A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur-containing natural gas based on ST UPFNN algorithms purify the dynamic evolutionary modeling method of technique
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
CN107436411A (en) * 2017-07-28 2017-12-05 南京航空航天大学 Battery SOH On-line Estimation methods based on fractional order neural network and dual-volume storage Kalman

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106777466A (en) * 2016-11-14 2017-05-31 重庆科技学院 High sulfur-containing natural gas based on ST UPFNN algorithms purify the dynamic evolutionary modeling method of technique
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
CN107436411A (en) * 2017-07-28 2017-12-05 南京航空航天大学 Battery SOH On-line Estimation methods based on fractional order neural network and dual-volume storage Kalman

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
A new smallest sigma set for the Unscented Transform and its applications on SLAM;Henrique M. Menegaz,et al;《IEEE》;20120301;正文第3172-3177页 *

Also Published As

Publication number Publication date
CN108509692A (en) 2018-09-07

Similar Documents

Publication Publication Date Title
CN106777866B (en) Energy-saving consumption-reducing-oriented modeling and optimizing method for high-sulfur-content natural gas purification process
CN106777465B (en) Dynamic evolution modeling and energy-saving optimization method for high-sulfur natural gas purification process
CN103345559B (en) The dynamic evolutionary modeling method of aluminium electrolysis process art electrolysis cells energy consumption
CN104636600B (en) High sulfur-containing natural gas purification Technology Modeling, optimization method based on extreme learning machine
CN104656441B (en) Gas purifying process modeling optimization method based on Unscented kalman neutral net
CN108509692B (en) A Modeling Method for High Sulfur Natural Gas Desulfurization Process Based on MiUKFNN Algorithm
CN103927412B (en) Instant learning debutanizing tower soft-measuring modeling method based on gauss hybrid models
US20180029900A1 (en) A Method for Effluent Total Nitrogen-based on a Recurrent Self-organizing RBF Neural Network
CN109543916B (en) A model for predicting the growth rate of silicon rods in a polysilicon reduction furnace
CN102562469B (en) Short-term wind driven generator output power predicting method based on correction algorithm
CN116822381B (en) An artificial intelligence-based ocean temperature and salt structure inversion method
CN104537415A (en) Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM
CN106227038A (en) Grain drying tower intelligent control method based on neutral net and fuzzy control
CN106503867A (en) A kind of genetic algorithm least square wind power forecasting method
CN109670625B (en) NOx emission concentration prediction method based on unscented Kalman filter least squares support vector machine
CN107273633B (en) Varying delay method of estimation and flow time lag method of estimation is hydrocracked between multiple operation
CN107169610A (en) Aquaculture dissolved oxygen prediction method and device
CN104865827B (en) Oil pumping unit oil extraction optimization method based on multi-working-condition model
CN106777466B (en) Dynamic evolution modeling method of high-sulfur natural gas purification process based on ST-UPFNN algorithm
CN105807609A (en) Multi-parameter mine spraying heat exchange efficiency calculation method-based spraying system
CN108595892A (en) Soft-measuring modeling method based on time difference model
CN105868559A (en) Atmospheric particulate mass concentration fitting method
CN111476402A (en) Wind power generation capacity prediction method coupling meteorological information and EMD technology
CN115829157A (en) Chemical water quality index prediction method based on variational modal decomposition and auto former model
CN113984989A (en) Aquaculture water quality abnormity detection method based on Laplace dimensionality reduction

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
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Guangzhou Senyu automation machinery design Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980040566

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230830

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Foshan shangxiaoyun Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041008

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230906

Application publication date: 20180907

Assignee: FOSHAN YAOYE TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041004

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230906

Application publication date: 20180907

Assignee: FOSHAN YIQING TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041003

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230906

Application publication date: 20180907

Assignee: Guangzhou Fuke Machinery Trade Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980040999

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230906

Application publication date: 20180907

Assignee: Guangzhou trump Environmental Protection Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980040995

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230906

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Wokang (Guangzhou) Electronic Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041471

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230913

Application publication date: 20180907

Assignee: Changhong (Guangzhou) Information Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041467

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230913

Application publication date: 20180907

Assignee: Dongguan Yaluo Environmental Protection Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041460

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230913

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Laishi (Guangzhou) Digital Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041991

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Guangzhou Qiming Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041990

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Guangzhou Daguan Digital Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041989

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Yichang Dae Urban and Rural Construction Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041988

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Guangzhou Dongtong Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041866

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Dongguan Zhaoyi Information Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041863

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

Application publication date: 20180907

Assignee: Leta (Guangzhou) Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980041859

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20230922

EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: GUANGZHOU KUAIZHOU INTELLIGENT ENVIRONMENTAL TECHNOLOGY CO.,LTD.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044603

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: Guangzhou Tuyu Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044600

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: GUANGZHOU SHANGCHENG TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044597

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: GUANGZHOU JUFENG TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044596

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: GUANGZHOU XINGYIN TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044593

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: GUANGZHOU LVNENG INTELLIGENT TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044591

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: Guangzhou Xiaoqing Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044587

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: Guangzhou Fangshao Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044586

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

Application publication date: 20180907

Assignee: Guangzhou star automation equipment Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980044559

Denomination of invention: A Modeling Method for High Sour Natural Gas Desulfurization Process Based on MiUKFNN Algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231031

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Guangzhou Yuming Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047712

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: Yajia (Guangzhou) Electronic Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047706

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: Guangzhou Yibo Yuntian Information Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047705

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: GUANGZHOU XIAONAN TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047703

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: GUANGZHOU YIDE INTELLIGENT TECHNOLOGY Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047702

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: Lingteng (Guangzhou) Electronic Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047701

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: Guangzhou Taipu Intelligent Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047700

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

Application publication date: 20180907

Assignee: Yuxin (Guangzhou) Electronic Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980047695

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231124

EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Guangxi GaoMin Technology Development Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2023980053986

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20231227

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Foshan chopsticks Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980003017

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240322

Application publication date: 20180907

Assignee: Foshan qianshun Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980003012

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240322

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Foshan helixing Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980004524

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240419

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20180907

Assignee: Shuzhi Nongcang (Shandong) Ecological Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980008105

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240701

Application publication date: 20180907

Assignee: Qishi (Yantai) data Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980008104

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240701

Application publication date: 20180907

Assignee: Yantai Xingyue coating equipment Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980008099

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240701

Application publication date: 20180907

Assignee: Yantai Zhonglan Environmental Technology Co.,Ltd.

Assignor: Chongqing University of Science & Technology

Contract record no.: X2024980008084

Denomination of invention: A modeling method for high sulfur natural gas desulfurization process based on MiUKFNN algorithm

Granted publication date: 20220325

License type: Common License

Record date: 20240701