WO2020199665A1 - 一种多目标原油调合在线优化方法 - Google Patents

一种多目标原油调合在线优化方法 Download PDF

Info

Publication number
WO2020199665A1
WO2020199665A1 PCT/CN2019/126926 CN2019126926W WO2020199665A1 WO 2020199665 A1 WO2020199665 A1 WO 2020199665A1 CN 2019126926 W CN2019126926 W CN 2019126926W WO 2020199665 A1 WO2020199665 A1 WO 2020199665A1
Authority
WO
WIPO (PCT)
Prior art keywords
blending
oil
component
individual
individuals
Prior art date
Application number
PCT/CN2019/126926
Other languages
English (en)
French (fr)
Inventor
钱锋
钟伟民
何仁初
杜文莉
Original Assignee
华东理工大学
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 华东理工大学 filed Critical 华东理工大学
Priority to JP2021557699A priority Critical patent/JP2022528657A/ja
Publication of WO2020199665A1 publication Critical patent/WO2020199665A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04Adaptive 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

Definitions

  • the invention relates to the field of crude oil processing in refineries, in particular to an optimization method for crude oil blending.
  • the invention aims to provide an online optimization method for multi-object crude oil blending.
  • the present invention provides an online optimization method for multi-target crude oil blending.
  • the method includes the following steps:
  • the blending task parameters include the component oils involved in blending, the corresponding storage tank number and blending tank number, the upper and lower limits of each blending component oil formula, and the blending head
  • the crude oil properties include density, sulfur content, acid value, and naphtha yield.
  • w c and w g respectively represent the lowest weight of blending cost and the smallest weight of deviation between crude oil density and target density after blending;
  • ⁇ m represents the target density of the blending head
  • P ⁇ represents the blending predicted value of the component oil density at the blending head
  • TP(P ⁇ ) represents the predicted value of crude oil properties in the blending tank, and is a function of the predicted blending density P ⁇ of the component oil at the blending head;
  • t represents the remaining time of blending in this batch
  • Bc i represents the maximum flow rate of the i-th component in the refining line
  • V i represents the reserve of component oil i
  • r i_min and r i_max respectively represent the upper and lower limits of the formula of component oil i;
  • s l_min and s l_max respectively represent the upper and lower limits of the attribute value l;
  • P l represents the blending prediction value of the component oil properties at the blending head
  • TP(P l ) represents the predicted value of the crude oil properties in the blending tank, and is a function of the predicted value P l of the blending properties of the component oil at the blending head.
  • formulas (2) and (3) are used to obtain the predicted values TP(P ⁇ ) and TP(P l ) of the crude oil density and attributes in the blending tank required to obtain the attribute index of the whole tank crude oil:
  • TVOL means the target quality of blending
  • HVOL means the quality of blended tank bottom oil
  • T ⁇ represents the oil density at the bottom of the tank
  • T l represents the value of the attribute of the bottom oil l.
  • non-dominated sort-based adaptive differential evolution Nondominated Sorting Adaptive Differential Evolution, NSJADE for short optimization algorithm is used to obtain the optimal formula:
  • randci represents normal distribution
  • randci represents Cauchy distribution
  • mean A represents ordinary arithmetic average
  • mean L represents ordinary Lehmer average, as shown in the following formula (4):
  • u j,i,g represent the j-th component of the test vector u i of the i-th individual in the g-th generation;
  • v j, i, g denotes the j-th component of the i-th g Generation individual variation of the vector V i;
  • x j,i,g represent the j-th component of the i-th individual x i in the g-th generation
  • j rank is randomly selected to ensure the introduction of mutation information.
  • the non-dominated ranking-based adaptive differential evolution optimization algorithm uses non-dominated ranking to evaluate and rank individuals, where Pareto dominance is used as the first indicator of individual evaluation, and crowded distance is used as the evaluation
  • the second indicator of, the steps are:
  • step (6) If there is no individual who is not classified into a certain level, go to step (6), otherwise go back to step (3);
  • Imq as the difference between two adjacent individuals in the m-th target in the ranking of the m-th target among the same-level individuals, which is a positive value.
  • Its Imq is positive infinity, and at this level, for the m-th target, its maximum and minimum values are f m_max and f m_min , and the crowding distance of each body q is calculated by formula (6):
  • the present invention provides an intelligent optimization algorithm to make it easier to find the global optimal solution of the problem, so as to better guide the production process.
  • Figure 1 is a structural diagram of the online optimization system for crude oil blending.
  • Figure 2 is a flowchart of an online optimization method for crude oil blending.
  • Figure 3 is a flowchart of an adaptive differential evolution optimization algorithm based on non-dominated sorting.
  • the inventor considered the lowest cost of blending, the smallest deviation between the density of crude oil after blending and the target density, and the constraints of various crude oil quality indicators (sulfur content, acid value, naphtha yield, nitrogen content, etc.).
  • Component oil inventory restrictions, formulation restrictions and flow restrictions of each component oil, etc. Under the condition of satisfying each constraint, the optimization goal is fully weighed, and the obtained optimal formula is sent to the control system to maximize the comprehensive benefits of actual production. On this basis, the present invention has been completed.
  • the multi-objective online optimization method for crude oil blending includes the following steps:
  • the parameters are initialized by blending
  • the second step is to set the optimization period, the weight of the objective function, the upper and lower limits of each attribute index of the blended crude oil, the reserves of each component oil, the maximum flow rate of each component oil in the refining line, and/or the unit quality cost of each component oil;
  • the third step is to obtain attribute data of each blended component oil and tank bottom oil according to the optimization cycle, and update the tank bottom dipstick, component oil storage, and the remaining time of the batch of blending;
  • the fourth step is to calculate the optimal formula
  • the fifth step is to send the current optimal formula to the blending control system for execution.
  • the component oil products participating in the blending are selected in the first step, the corresponding storage tank number and the blending tank number, the upper and lower limits of each blending component oil formula, Blending head flow, tank bottom oil dipstick, target blending dipstick, target blending density value, and/or blending time of this batch.
  • the system may include at least two crude oil component tanks, and the blending process is performed in batches, and the blending amount of each batch is known and determined by the dispatching department according to the production situation.
  • the fourth step described above adopts an adaptive differential evolution optimization algorithm based on non-dominated sorting to obtain the optimal formula of each blending component of the current optimization cycle under certain constraint conditions.
  • the goal is to achieve the lowest cost and the smallest deviation between the crude oil density and the target density after blending under certain constraints.
  • the optimal blending formula is obtained by solving formulas (1), (2), (3):
  • i (i 1, 2,..., n) blending component oil product label
  • r i the formula value of each component oil to be optimized
  • c i represents the unit quality cost of each component oil
  • w c , w g represent the lowest weight of blending cost, and the deviation between the crude oil density and the target density after blending is the smallest Weight
  • ⁇ m represents the target density of the blending
  • P ⁇ represents the blending predicted value of the component oil density at the blending head
  • TP(P ⁇ ) represents the predicted value of the crude oil properties in the blending tank, which is the blending head
  • t represents the remaining blending time of the batch;
  • Bc i represents the maximum flow rate of the i-th component refining line;
  • B represents the flow rate
  • the predicted values P ⁇ and P l of the blending density and properties of the component oil at the blending head are calculated by the principle of linear superposition, and then the density and properties obtained by the optimized blending are used to compensate the density and properties of the blended bottom oil Deviation, so that the density of the whole blending tank reaches the target value and the properties are qualified.
  • the predicted values TP(P ⁇ ) and TP(P l ) of the crude oil density and attributes in the blending tank can be calculated according to the following formula (2)(3) according to the requirements of the whole tank crude oil attribute index:
  • TVOL represents the target quality of blending
  • HVOL represents the quality of blended tank bottom oil
  • T ⁇ represents the density of tank bottom oil
  • T l represents the value of the attribute of tank bottom oil
  • randci represents normal distribution
  • randci represents Cauchy distribution
  • mean A represents ordinary arithmetic average
  • mean L represents ordinary Lehmer average, as shown in the following formula (4):
  • u j,i,g represent the j-th component of the test vector u i of the i-th individual in the g-th generation
  • v j,i,g represent the j-th component of the variation vector v i of the i-th individual in the g-th generation Component
  • x j, i, g represents the j-th component of the i-th individual x i in the g-th generation
  • j rank is a randomly selected j to ensure the introduction of mutation information.
  • the non-dominant ranking method used to evaluate and rank individuals is based on Pareto dominance as the first indicator of individual evaluation and crowding distance as the second indicator of evaluation. The steps are as follows:
  • step (6) If there is no individual who is not classified into a certain level, go to step (6), otherwise go back to step (3).
  • Imq as the difference between two adjacent individuals in the m-th target in the ranking of the m-th target among the same-level individuals, which is a positive value.
  • Its Imq is positive infinity
  • the maximum and minimum values of the m-th target at this level are f m_max and f m_min , and the crowding distance of each body q is calculated according to the following formula (6):
  • it further includes a sixth step, judging whether the blending is completed, if it is completed, stop here; if it is not completed, wait for an optimization cycle, and return to the third step.
  • the multi-objective online optimization method for crude oil blending proposed in the present invention adopts an adaptive differential evolution optimization method based on non-dominated sorting for optimization.
  • the optimization method provided by the present invention takes into account the lowest economic cost while considering the minimum deviation between the crude oil density and the target density after blending.
  • the two optimization goals are fully weighed using the adaptive differential algorithm based on non-dominated sorting, thereby achieving The overall benefit is the best.
  • the workflow of the crude oil blending optimization system mainly includes the following steps:
  • Step 1 Initialize the coordination task parameters.
  • the selected blended component oils are 1#, 2#, and 3# component oils, and the corresponding storage tank number and blending tank number are selected.
  • Step 2 Set the optimization period, the weight of the objective function, the upper and lower limits of optimization, the reserves of each component oil, and the unit quality cost of each component oil.
  • the lower limit of sulfur content in the blending tank is 0% and the upper limit is 2.5%; the lower limit of acid value is 0mg KOH and the upper limit is 0.5mg KOH; the lower limit of naphtha yield is 0.19% and the upper limit is 0.21%.
  • the input inventory of the three components is 4000t, 7200t and 4500t respectively.
  • the maximum flow rate of the component oil of each participating line and the unit mass cost of each component oil are shown in Table 1 below.
  • Step 3 Obtain the attribute data of each blended component oil and tank bottom oil according to the optimization cycle, and update the tank bottom dipstick, component oil storage, and the remaining blending time of the batch.
  • Step 4 Calculate the optimal formula.
  • the optimal formula obtained is as follows.
  • the 1# component oil formula corresponding to the blending head 1 is 0.23399
  • the 2# component oil formula is 0.36726
  • the 3# component oil formula is 0.39876.
  • the cost is 85.8308 US dollars/barrel.
  • the predicted blending The final density of the oil in the tank is 830.1294kg/m 3
  • the final properties of the oil in the blending tank are 2.0501% sulfur content, acidity 0.36851 mg KOH, and naphtha yield 0.19945%.
  • Step 5 Send the current optimal formula to the blending control system for execution.
  • Step 6 Determine whether the current blending is completed, if the current blending is not completed, wait for an optimization cycle, and then return to step 3.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Production Of Liquid Hydrocarbon Mixture For Refining Petroleum (AREA)
  • Feedback Control In General (AREA)

Abstract

本发明公开了一种多目标原油调合在线优化方法。所述方法包括步骤:首先,进行调合任务参数初始化;其次,设置优化周期、目标函数权重,设置调合原油各属性指标上下限、各组分油储量、各参炼线组分油最大流量以及各组分油的单位质量成本;再次,根据预定优化周期获取各调合组分及罐底油属性数据,更新罐底油尺、组分油储量、本批次的调合剩余时间;和最后,求取当前优化周期各调合组分的最优配方,并送至调合控制系统执行。

Description

一种多目标原油调合在线优化方法 技术领域
本发明涉及炼油企业原油加工领域,尤其是原油调合的优化方法。
背景技术
为实现综合效益最大化,在原油调合过程中往往需要对多种优化目标进行权衡。而在原油调合实际工业中,由于传统算法计算能力的限制,炼油企业大多仅能进行单目标优化,而忽略了其他目标,从而无法做到充分的权衡。如在要求调合成本最低时,就无法将密度等属性控制到某个目标值,只要求在一定范围内,牺牲了生产过程的可控性,限制了生产水平以及产品质量的进一步提高。而在对密度等属性提出要求时,由于缺少成本最小的目标,失去了对方案的比较环节,造成了不必要的浪费。
因此,本领域迫切需要提供一种智能优化算法,以便更容易找到问题的全局最优解,从而更好地指导生产过程。
发明内容
本发明旨在提供一种多目标原油调合在线优化方法。
本发明提供一种多目标原油调合在线优化方法,所述方法包括以下步骤:
首先,进行调合任务参数初始化;
其次,设置优化周期、目标函数权重,设置调合原油各属性指标上下限、各组分油储量、各参炼线组分油最大流量以及各组分油的单位质量成本;
再次,根据预定优化周期获取各调合组分及罐底油属性数据,更新罐底油尺、组分油储量、本批次的调合剩余时间;和
最后,求取当前优化周期各调合组分的最优配方,并送至调合控制系统执行。
在另一优选例中,所述调合任务参数包括参与调合的组分油品、相应储罐罐号以及调合罐罐号、各调合组分油配方的上限与下限、调合头流量、目标调合油尺、目标调合密度值以及本批次调合时间;更佳地,所述调合任务参数还包括罐底油油尺。
在另一优选例中,所述原油属性包括密度、硫含量、酸值和石脑油收率。
在另一优选例中,建立了如公式(1)所示的原油调合在线优化多目标数学模型:
Figure PCTCN2019126926-appb-000001
其中,
i(i=1,2,…,n)调合组分油品标号;
l(l=1,2,…,n)表示原油属性指标(硫含量、酸值、石脑油收率等);
r i待优化的各组分油配方值;c i表示各组分油的单位质量成本;
w c,w g分别表示调合成本最低权值,和调合后原油密度与目标密度偏差最小权值;
ρ m表示调合头的目标密度;
P ρ表示调合头处组分油密度的调合预测值;
TP(P ρ)表示调合罐中的原油属性预测值,是调合头处组分油的调合密度预测值P ρ的函数;
t表示本批次的调合剩余时间;
Bc i表示第i个组分参炼线最大流量;
B表示调合头本批次的流量;
V i表示组分油i的储量;r i_min和r i_max分别表示组分油i的配方的上下限;
s l_min和s l_max分别表示属性值l的上下限;
P l表示调合头处组分油属性的调合预测值;
TP(P l)表示调合罐中的原油属性预测值,是调合头处组分油的调合属性预测值P l的函数。
在另一优选例中,采用式(2)和式(3)得到全罐原油属性指标要求调合罐中的原油密度与属性的预测值TP(P ρ)与TP(P l):
Figure PCTCN2019126926-appb-000002
Figure PCTCN2019126926-appb-000003
其中,
TVOL表示调合目标质量;
HVOL表示调合罐底油质量;
T ρ表示罐底油密度;
T l表示罐底油l属性的值。
在另一优选例中,采用基于非支配排序的自适应差分进化(Nondominated Sorting Adaptive Differential Evolution,简称NSJADE,)优化算法求取最优配方:
(1)初始化种群P,种群大小为NP;
(2)初始化μ CR=0.5,μ F=0.5,
Figure PCTCN2019126926-appb-000004
(3)对种群P进行非支配排序;
(4)循环开始,当算法终止条件尚未满足时,进行:
(i)设置
Figure PCTCN2019126926-appb-000005
(ii)针对种群P中的每一个个体x i,生成对应的比例因子F i=randc iF,0.1)和交叉概率CR i=randn iCR,0.1);
(iii)从适应度值前100p%的个体中随机挑选一个个体,记为x p,best,从种群P中选择个体x r1,x r1≠x i,从种群P∪A中选择个体x r2,x r2≠x r1≠x i
(iv)生成变异向量v i=x i+F i·(x p,best-x i)+F i·(x r1-x r2);
(v)生成试验向量u i
(vi)比较x i和u i的适应度函数值,若u i优于x i,则将x i放入A中,F i放入S F中,CR i放入S CR中;
(vii)在每一代更新结束后,随机移除A中的个体,使得|A|≤NP;
(viii)更新μ F和μ CR:μ F=(1-c)·μ F+c·mean L(S F),μ CR=(1-c)·μ CR+c·mean A(S CR)
(ix)合并父代种群x和子代种群u,进行非支配排序操作,筛选前NP个个体为下一代的种群;
(5)算法停止,得到最终种群NP;
其中,randci表示正态分布;randci表示柯西分布;mean A表示普通的算术平均;mean L表示普通的Lehmer平均,如下式(4)所示:
Figure PCTCN2019126926-appb-000006
S F与S CR分别用于储存成功产生相比于父代x i更优秀试验向量u i的F i与CR i; 产生试验向量的方法如以下式(5)所示:
Figure PCTCN2019126926-appb-000007
其中,
u j,i,g表示第g代第i个个体的试验向量u i的第j个分量;
v j,i,g表示第g代第i个个体的变异向量v i的第j个分量;
x j,i,g表示第g代第i个个体x i的第j个分量;
j rank为保证变异信息的引入而随机选定的一个j。
在另一优选例中,所述基于非支配排序的自适应差分进化优化算法,采用非支配排序对个体进行评价与排序,其中,Pareto支配关系作为个体评价的第一个指标,拥挤距离作为评价的第二个指标,步骤为:
①对于各个体p,创建支配该个体的个体集Sp,并设置Sp中个体数为np,设置i=1;
②对于np为0的个体将其归于第1级,该个体其所在级数p rank=1,i=i+1;
③将属于第1级的个体从含该个体的Sp中删去,相应np减1;
④对于np为0的个体将其归于第i级,该个体其所在级数p rank=i,i=i+1;
⑤若不存在未归入某一级中的个体则转到第(6)步,否则返回第(3)步;
⑥对于同级的个体,对于各目标函数分别进行一次排序;
⑦对于个体q,定义I.m.q为其在同级个体中在关于第m个目标的排序中相邻的两个个体在第m个目标的差值,其为正值,对于在边界上的个体,其I.m.q为正无穷,并设在该级上关于第m个目标,其最大值与最小值为f m_max与f m_min,采用式(6)分别计算各个体q的拥挤距离:
Figure PCTCN2019126926-appb-000008
⑧按以下准则进行排序,对于个体i与j,规定(a)i rank<j rank时或(b)i rank=j rank且i distance>j distance时,个体i优秀于j,将所有个体从优到劣进行排序。
据此,本发明提供了一种智能优化算法,以便更容易找到问题的全局最优解,从而更好地指导生产过程。
附图说明
图1是原油调合在线优化系统结构图。
图2是原油调合在线优化方法流程图。
图3是基于非支配排序的自适应差分进化优化算法流程图。
具体实施方式
发明人经过广泛而深入的研究,针对炼油企业原油调合生产中由于传统算法的能力限制仅能进行单目标优化,无法进行各优化目标的充分权衡的问题,提出了一种原油调合的多目标在线优化方法,并采用基于非支配排序的自适应差分进化优化方法进行优化求解。
例如,发明人考虑了调合成本最低、调合后原油密度与目标密度偏差最小、同时考虑了各种原油质量指标约束(硫含量、酸值、石脑油收率、氮含量等)、各组分油的库存限制、各组分油的配方限制和流量限制等。在满足各约束的条件下,实现对优化目标的充分权衡,并将所得最优配方送入控制系统,实现实际生产的综合效益的最大化。在此基础上,完成了本发明。
具体地,本发明提供的原油调合的多目标在线优化方法包括以下步骤:
第一步,进行调合认为参数初始化;
第二步,设置优化周期、目标函数权重、设置调合原油各属性指标上下限、各组分油储量、各参炼线组分油最大流量、和/或各组分油的单位质量成本;
第三步,根据优化周期获取各调合组分油以及罐底油的属性数据,更新罐底油尺、组分油储量、本批次的调合剩余时间;
第四步,计算最优配方;
第五步,将当前最优配方送至调合控制系统执行。
在本发明的一种实施方式中,上述第一步中选定参与调合的组分油品、相应储罐罐号以及调合罐罐号、各调合组分油配方的上限与下限、调合头流量、罐底油油尺、目标调合油尺、目标调合密度值、和/或本批次调合时间。
在本发明的一个实施例中,系统中可包含至少两个原油组分罐,调合过程按批次进行,每批次调合量是已知的,由调度部门根据生产情况确定。
在本发明的一种实施方式中,上述第四步在满足一定约束条件下,采用基于非支配排序的自适应差分进化优化算法求取当前优化周期各调合组分的最优配方。
例如但不限于,在满足一定约束条件下实现成本最低以及调合后原油密度与目标密度偏差最小作为目标。对于单个优化周期内,通过对公式(1)、(2)、(3)求解得到最 优调合配方:
Figure PCTCN2019126926-appb-000009
其中,i(i=1,2,…,n)调合组分油品标号;l(l=1,2,…,n)表示原油属性指标(硫含量、酸值、石脑油收率等);r i待优化的各组分油配方值;c i表示各组分油的单位质量成本;w c,w g表示调合成本最低权值,调合后原油密度与目标密度偏差最小权值;ρ m表示调合的目标密度;P ρ表示调合头处组分油密度的调合预测值;TP(P ρ)表示表示调合罐中的原油属性预测值,是调合头处组分油的调合密度预测值P ρ的函数;t表示本批次的剩余调合时间;Bc i表示第i个组分参炼线最大流量;B表示调合头本批次的流量;V i表示组分油i的储量;r i_min和r i_max分别表示组分油i的配方的上下限;s l_min和s l_max分别表示属性值l的上下限;P l表示调合头处组分油属性的调合预测值;TP(P l)表示调合罐中的原油属性预测值,是调合头处组分油的调合属性预测值P l的函数。
调合头处组分油的调合密度与属性的预测值P ρ与P l以线性叠加原理进行计算,再利用优化调合所得的密度以及属性补偿已调合的罐底油的密度与属性偏差,使整个调合罐密度达到目标值且属性合格。
根据全罐原油属性指标要求调合罐中的原油密度与属性的预测值TP(P ρ)与TP(P l)可以按照下式(2)(3)进行计算:
Figure PCTCN2019126926-appb-000010
Figure PCTCN2019126926-appb-000011
其中,TVOL表示调合目标质量;HVOL表示调合罐底油质量;T ρ表示罐底油密度;T l表示罐底油l属性的值;
然后采用基于非支配排序的自适应差分进化优化算法进行配方优化,求解以上多目标优化问题,得出优化调合配方。步骤如下:
1)初始化种群P,种群大小为NP;
2)初始化μ CR=0.5,μ F=0.5,
Figure PCTCN2019126926-appb-000012
3)对种群P进行非支配排序;
4)循环开始,当算法终止条件尚未满足时,进行:
i)设置
Figure PCTCN2019126926-appb-000013
ii)针对种群P中的每一个个体x i,生成对应的比例因子F i=randc iF,0.1)和交叉概率CR i=randn iCR,0.1);
iii)从适应度值前100p%的个体中随机挑选一个个体,记为x p,best,从种群P中选择个体x r1,x r1≠x i,从种群P∪A中选择个体x r2,x r2≠x r1≠x i
iv)生成变异向量v i=x i+F i·(x p,best-x i)+F i·(x r1-x r2);
v)生成试验向量u i
vi)比较x i和u i的适应度函数值,若u i优于x i,则将x i放入A中,F i放入S F中,CR i放入S CR中;
vii)在每一代更新结束后,随机移除A中的个体,使得|A|≤NP;
viii)更新μ F和μ CR:μ F=(1-c)·μ F+c·mean L(S F),μ CR=(1-c)·μ CR+c·mean A(S CR)
ix)合并父代种群x和子代种群u,进行非支配排序操作,筛选前NP个个体为下一代的种群;
5)算法停止,得到最终种群NP。
其中,randci表示正态分布;randci表示柯西分布;mean A表示普通的算术平均;mean L表示普通的Lehmer平均,如下式(4)所示:
Figure PCTCN2019126926-appb-000014
S F与S CR分别用于储存成功产生相比于父代x i更优秀试验向量u i的F i与CR i;产生试验向量的方法如以下式(5)所示:
Figure PCTCN2019126926-appb-000015
其中,u j,i,g表示第g代第i个个体的试验向量u i的第j个分量;v j,i,g表示第g代第i个个体的变异向量v i的第j个分量;x j,i,g表示第g代第i个个体x i的第j个分量;j rank为保证变异信息的引入而随机选定的一个j。
其中采用的非支配排序对个体进行评价与排序的方法,是以Pareto支配关系作为个体评价的第一个指标,以拥挤距离作为评价的第二个指标进行排序的,步 骤为:
①对于各个体p,创建支配该个体的个体集Sp,并设置Sp中个体数为np,设置i=1。
②对于np为0的个体将其归于第1级,该个体其所在级数p rank=1,i=i+1。
③将属于第1级的个体从含该个体的Sp中删去,相应np减1。
④对于np为0的个体将其归于第i级,该个体其所在级数p rank=i,i=i+1。
⑤若不存在未归入某一级中的个体则转到第(6)步,否则返回第(3)步。
⑥对于同级的个体,对于各目标函数分别进行一次排序
⑦对于个体q,定义I.m.q为其在同级个体中在关于第m个目标的排序中相邻的两个个体在第m个目标的差值,其为正值,对于在边界上的个体,其I.m.q为正无穷,并设在该级上关于第m个目标,其最大值与最小值为f m_max与f m_min,按以下式(6)分别计算各个体q的拥挤距离:
Figure PCTCN2019126926-appb-000016
⑧按以下准则进行排序,对于个体i与j,规定(i)i rank<j rank时或(ii)i rank=j rank且i distance>j distance时,个体i优秀于j,将所有个体从优到劣进行排序。
在本发明的一种实施方式中,还包括第六步,判断调合是否完成,若完成,则止于此;若未完成,则等待一个优化周期,返回第三步。
本发明提到的上述特征,或实施例提到的特征可以任意组合。本案说明书所揭示的所有特征可与任何组合物形式并用,说明书中所揭示的各个特征,可以任何可提供相同、均等或相似目的的替代性特征取代。因此除有特别说明,所揭示的特征仅为均等或相似特征的一般性例子。
本发明的主要优点在于:
1、本发明提出的原油调合的多目标在线优化方法,采用基于非支配排序的自适应差分进化优化方法进行优化求解。
2、本发明提供的优化方法在考虑经济成本最低的同时考虑了调合后原油密度与目标密度偏差最小,利用基于非支配排序的自适应差分算法对两个优化目标进行充分权衡,从而实现了总体效益最优。
下面结合具体实施例,进一步阐述本发明。应理解,这些实施例仅用于说明本 发明而不用于限制本发明的范围。下列实施例中未注明具体条件的实验方法,通常按照常规条件或按照制造厂商所建议的条件。除非另外说明,否则所有的百分数、比率、比例、或份数按重量计。除非另行定义,文中所使用的所有专业与科学用语与本领域熟练人员所熟悉的意义相同。此外,任何与所记载内容相似或均等的方法及材料皆可应用于本发明方法中。文中所述的较佳实施方法与材料仅作示范之用。
下面以本发明在某原油调合过程的实际实施情况并结合一具体算例,给出详细的计算过程与操作流程。实施例所涉及的系统结构如图1所示,该系统包含三个原油组分罐,调合过程按照批次进行,每批次的调合量是已知的,由调度部门根据生产情况确定,一般在8000t-50000t,下面给出详细的计算过程和具体操作流程。本算例在以本发明技术方案为前提进行实施。
如图2所示,原油调合优化系统的工作流程主要包括以下步骤:
步骤一:调合任务参数初始化。
结合具体算例,选定调合的组分油品为1#、2#以及3#组分油,并选择相应储罐罐号以及调合罐罐号。
设定1#组分油品配方下限为0,配方上限0.35;2#组分油品配方下限为0.2,配方上限0.6;3#组分油品配方下限为0,配方上限0.4。设定调合头的调合流量为1000t/h,本批次调合时间为10h,本批次的调合目标为11000t,目标调合密度值为830kg/m 3
步骤二:设置优化周期、目标函数权重、优化上下限、各组分油储量以及各组分油的单位质量成本。
设置优化假设为5min,并设置调合成本最低权值为0.4,调合后原油密度与目标密度偏差最小权值0.6。
选定调合后调合罐内硫含量下限为0%,上限2.5%;酸值下限为0mg KOH,上限为0.5mg KOH;石脑油收率下限为0.19%,上限为0.21%。输入三种组分的库存分别为4000t、7200t以及4500t。各参炼线组分油最大流量及各组分油的单位质量成本如下表1所示。
步骤三:根据优化周期获取各调合组分油以及罐底油的属性数据,更新罐底油尺、组分油储量、本批次的调合剩余时间。
获取的各调合组分油以及罐底油的属性数据如下表1所示:
表1
Figure PCTCN2019126926-appb-000017
更新罐底油尺、组分油储量,所得罐底油质量1000t,组分油储量分别为4000t、7200t以及4500t。本批次的调合剩余时间为10h。
步骤四:计算最优配方。
通过以上数据按公式(1)、(2)、(3)建立完模型后,调用基于非支配排序的自适应差分进化优化算法求解程序求解,求解过程如图3所示。
所得结果最优配方如下,调合头1对应的1#组分油配方为0.23399,2#组分油配方0.36726,3#组分油配方0.39876,成本为85.8308美元/桶,预测所得的调合罐中油品最终密度为830.1294kg/m 3,其调合罐中油品最终属性为硫含量2.0501%,酸度为0.36851mg KOH,石脑油收率为0.19945%。
步骤五:将当前最优配方送至调合控制系统执行。
步骤六:判断本次调合是否完成,若本次调合未完成,则进行等待一个优化周期,再返回步骤三。
以上所述仅为本发明的较佳实施例而已,并非用以限定本发明的实质技术内容范围,本发明的实质技术内容是广义地定义于申请的权利要求范围中,任何他人完成的技术实体或方法,若是与申请的权利要求范围所定义的完全相同,也或是一种等效的变更,均将被视为涵盖于该权利要求范围之中。

Claims (8)

  1. 一种多目标原油调合在线优化方法,其特征在于,所述方法包括以下步骤:
    首先,进行调合任务参数初始化;
    其次,设置优化周期、目标函数权重,设置调合原油各属性指标上下限、各组分油储量、各参炼线组分油最大流量以及各组分油的单位质量成本;
    再次,根据预定优化周期获取各调合组分及罐底油属性数据,更新罐底油尺、组分油储量、本批次的调合剩余时间;和
    最后,求取当前优化周期各调合组分的最优配方,并送至调合控制系统执行。
  2. 如权利要求1所述的方法,其特征在于,所述调合任务参数包括参与调合的组分油品、相应储罐罐号以及调合罐罐号、各调合组分油配方的上限与下限、调合头流量、目标调合油尺、目标调合密度值以及本批次调合时间。
  3. 如权利要求2所述的方法,其特征在于,所述调合任务参数还包括罐底油油尺。
  4. 如权利要求1所述的方法,其特征在于,所述原油属性包括密度、硫含量、酸值和石脑油收率。
  5. 如权利要求1所述的方法,其特征在于,建立了如公式(1)所示的原油调合在线优化多目标数学模型:
    Figure PCTCN2019126926-appb-100001
    其中,
    i(i=1,2,…,n)调合组分油品标号;
    l(l=1,2,…,n)表示原油属性指标(硫含量、酸值、石脑油收率等);
    r i待优化的各组分油配方值;
    c i表示各组分油的单位质量成本;
    w c,w g分别表示调合成本最低权值,和调合后原油密度与目标密度偏差最小权值;
    ρ m表示调合头的目标密度;
    P ρ表示调合头处组分油密度的调合预测值;
    TP(P ρ)表示调合罐中的原油属性预测值,是调合头处组分油的调合密度预测值P ρ的函数;
    t表示本批次的调合剩余时间;
    Bc i表示第i个组分参炼线最大流量;
    B表示调合头本批次的流量;
    V i表示组分油i的储量;
    r i_min和r i_max分别表示组分油i的配方的上下限;s l_min和s l_max分别表示属性值l的上下限;
    P l表示调合头处组分油属性的调合预测值;
    TP(P l)表示调合罐中的原油属性预测值,是调合头处组分油的调合属性预测值P l的函数。
  6. 如权利要求5所述的方法,其特征在于,采用式(2)和式(3)得到全罐原油属性指标要求调合罐中的原油密度与属性的预测值TP(P ρ)与TP(P l):
    Figure PCTCN2019126926-appb-100002
    Figure PCTCN2019126926-appb-100003
    其中,
    TVOL表示调合目标质量;
    HVOL表示调合罐底油质量;
    T ρ表示罐底油密度;
    T l表示罐底油l属性的值。
  7. 如权利要求1所述的方法,其特征在于,采用基于非支配排序的自适应差分进化(Nondominated Sorting Adaptive Differential Evolution,简称NSJADE,)优化算法求取最优配方:
    (1)初始化种群P,种群大小为NP;
    (2)初始化μ CR=0.5,μ F=0.5,
    Figure PCTCN2019126926-appb-100004
    (3)对种群P进行非支配排序;
    (4)循环开始,当算法终止条件尚未满足时,进行:
    (i)设置
    Figure PCTCN2019126926-appb-100005
    (ii)针对种群P中的每一个个体x i,生成对应的比例因子F i=randc iF,0.1)和交叉概率CR i=randn iCR,0.1);
    (iii)从适应度值前100p%的个体中随机挑选一个个体,记为x p,best,从种群P中选择个体x r1,x r1≠x i,从种群P∪A中选择个体x r2,x r2≠x r1≠x i
    (iv)生成变异向量v i=x i+F i·(x p,best-x i)+F i·(x r1-x r2);
    (v)生成试验向量u i
    (vi)比较x i和u i的适应度函数值,若u i优于x i,则将x i放入A中,F i放入S F中,CR i放入S CR中;
    (vii)在每一代更新结束后,随机移除A中的个体,使得|A|≤NP;
    (viii)更新μ F和μ CR:μ F=(1-c)·μ F+c·mean L(S F),μ CR=(1-c)·μ CR+c·mean A(S CR)
    (ix)合并父代种群x和子代种群u,进行非支配排序操作,筛选前NP个个体为下一代的种群;
    (5)算法停止,得到最终种群NP;
    其中,randci表示正态分布;randci表示柯西分布;mean A表示普通的算术平均;mean L表示普通的Lehmer平均,如下式(4)所示:
    Figure PCTCN2019126926-appb-100006
    S F与S CR分别用于储存成功产生相比于父代x i更优秀试验向量u i的F i与CR i;产生试验向量的方法如以下式(5)所示:
    Figure PCTCN2019126926-appb-100007
    其中,
    u j,i,g表示第g代第i个个体的试验向量u i的第j个分量;
    v j,i,g表示第g代第i个个体的变异向量v i的第j个分量;
    x j,i,g表示第g代第i个个体x i的第j个分量;
    j rank为保证变异信息的引入而随机选定的一个j。
  8. 如权利要求7所述的方法,其特征在于,所述基于非支配排序的自适应差分进化优化算法,采用非支配排序对个体进行评价与排序,其中,Pareto支配关系作为个体评价的第一个指标,拥挤距离作为评价的第二个指标,步骤为:
    ①对于各个体p,创建支配该个体的个体集Sp,并设置Sp中个体数为np,设置i=1;
    ②对于np为0的个体将其归于第1级,该个体其所在级数p rank=1,i=i+1;
    ③将属于第1级的个体从含该个体的Sp中删去,相应np减1;
    ④对于np为0的个体将其归于第i级,该个体其所在级数p rank=i,i=i+1;
    ⑤若不存在未归入某一级中的个体则转到第(6)步,否则返回第(3)步;
    ⑥对于同级的个体,对于各目标函数分别进行一次排序;
    ⑦对于个体q,定义I.m.q为其在同级个体中在关于第m个目标的排序中相邻的两个个体在第m个目标的差值,其为正值,对于在边界上的个体,其I.m.q为正无穷,并设在该级上关于第m个目标,其最大值与最小值为f m_max与f m_min,采用式(6)分别计算各个体q的拥挤距离:
    Figure PCTCN2019126926-appb-100008
    ⑧按以下准则进行排序,对于个体i与j,规定(a)i rank<j rank时或(b)i rank=j rank且i distance>j distance时,个体i优秀于j,将所有个体从优到劣进行排序。
PCT/CN2019/126926 2019-03-29 2019-12-20 一种多目标原油调合在线优化方法 WO2020199665A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2021557699A JP2022528657A (ja) 2019-03-29 2019-12-20 多目的原油混合用オンライン最適化方法

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910250894.2A CN109765788B (zh) 2019-03-29 2019-03-29 一种多目标原油调合在线优化方法
CN201910250894.2 2019-03-29

Publications (1)

Publication Number Publication Date
WO2020199665A1 true WO2020199665A1 (zh) 2020-10-08

Family

ID=66460206

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/126926 WO2020199665A1 (zh) 2019-03-29 2019-12-20 一种多目标原油调合在线优化方法

Country Status (3)

Country Link
JP (1) JP2022528657A (zh)
CN (1) CN109765788B (zh)
WO (1) WO2020199665A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112947343A (zh) * 2021-03-03 2021-06-11 中国石油化工股份有限公司 炼油化工全厂生产目标跟踪控制方法和装置
CN112966805A (zh) * 2021-03-15 2021-06-15 河海大学 一种基于图卷积神经网络和nsga-ⅱ算法的水库调度多目标优化方法
CN113110336A (zh) * 2021-04-20 2021-07-13 南京富岛信息工程有限公司 一种考虑调度约束的原油动态调合方法
CN113406274A (zh) * 2021-05-28 2021-09-17 中国人民解放军空军研究院航空兵研究所 基于粘度在线检测的混合均匀度分析方法及混合装置

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109765788B (zh) * 2019-03-29 2022-03-04 华东理工大学 一种多目标原油调合在线优化方法
CN111120873B (zh) * 2019-12-06 2021-11-30 浙江中控技术股份有限公司 用于油品在线调合的管道补偿方法
CN116224776A (zh) * 2022-11-03 2023-06-06 上海流程智造科技创新研究院有限公司 一种基于约束多目标的汽油调合在线优化方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103955778A (zh) * 2014-05-15 2014-07-30 南京富岛信息工程有限公司 一种高性能汽油调合非线性优化方法
CN105279581A (zh) * 2015-11-16 2016-01-27 电子科技大学 基于差分进化的geo-uav双基sar路径规划方法
CN106647262A (zh) * 2016-11-28 2017-05-10 中国人民解放军国防科学技术大学 一种面向敏捷卫星多目标任务规划的差分进化方法
CN108333943A (zh) * 2018-02-26 2018-07-27 华东理工大学 基于增量模式的原油调合在线优化方法
CN109765789A (zh) * 2019-03-29 2019-05-17 华东理工大学 考虑罐底油属性的原油调合双调合头协调优化方法
CN109765788A (zh) * 2019-03-29 2019-05-17 华东理工大学 一种多目标原油调合在线优化方法

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6816822B1 (en) * 2000-08-16 2004-11-09 Abb Automation Inc. System and method for dynamic modeling, parameter estimation and optimization for processes having operating targets
JP2003015728A (ja) * 2001-04-03 2003-01-17 Nippon Steel Corp 原料の最適混合演算方法、装置、コンピュータプログラム、及びコンピュータ読み取り可能な記憶媒体
JP3683189B2 (ja) * 2001-06-22 2005-08-17 住友軽金属工業株式会社 原料配合計画システム及び原料配合計画方法
CN101286065B (zh) * 2008-05-23 2010-06-02 清华大学 一种原油掺炼多周期优化的原油调合调度方法
US20110098862A1 (en) * 2009-10-27 2011-04-28 ExxonMobil Research Engineering Company Law Department Multi-stage processes and control thereof
CN102643662B (zh) * 2012-04-25 2014-09-17 南京富岛信息工程有限公司 一种原油调合优化方法
CN103365208A (zh) * 2013-07-12 2013-10-23 南京富岛信息工程有限公司 一种具有底油性质偏差修正的原油调合优化方法
CN103365205B (zh) * 2013-07-25 2016-01-27 南京富岛信息工程有限公司 一种原油调合双调合头在线协调控制方法
US10040041B2 (en) * 2015-07-09 2018-08-07 Cameron International Corporation Crude oil blending using total boiling point analysis
CN105809297B (zh) * 2016-05-18 2019-06-25 西南石油大学 一种基于多目标差分进化算法的火电厂环境经济调度方法
CN107480885A (zh) * 2017-08-14 2017-12-15 国家电网公司 基于非支配排序差分进化算法的分布式电源布点规划方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103955778A (zh) * 2014-05-15 2014-07-30 南京富岛信息工程有限公司 一种高性能汽油调合非线性优化方法
CN105279581A (zh) * 2015-11-16 2016-01-27 电子科技大学 基于差分进化的geo-uav双基sar路径规划方法
CN106647262A (zh) * 2016-11-28 2017-05-10 中国人民解放军国防科学技术大学 一种面向敏捷卫星多目标任务规划的差分进化方法
CN108333943A (zh) * 2018-02-26 2018-07-27 华东理工大学 基于增量模式的原油调合在线优化方法
CN109765789A (zh) * 2019-03-29 2019-05-17 华东理工大学 考虑罐底油属性的原油调合双调合头协调优化方法
CN109765788A (zh) * 2019-03-29 2019-05-17 华东理工大学 一种多目标原油调合在线优化方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WEI DU ET AL.: "Robust Order Scheduling in the Discrete Manufacturing Industry: A Multiobjective Optimization Approach", IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, vol. 14, no. 1, 31 January 2018 (2018-01-31), XP011675662, ISSN: 1551-3203, DOI: 20200225140038Y *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112947343A (zh) * 2021-03-03 2021-06-11 中国石油化工股份有限公司 炼油化工全厂生产目标跟踪控制方法和装置
CN112966805A (zh) * 2021-03-15 2021-06-15 河海大学 一种基于图卷积神经网络和nsga-ⅱ算法的水库调度多目标优化方法
CN112966805B (zh) * 2021-03-15 2022-10-18 河海大学 一种基于图卷积神经网络和nsga-ⅱ算法的水库调度多目标优化方法
CN113110336A (zh) * 2021-04-20 2021-07-13 南京富岛信息工程有限公司 一种考虑调度约束的原油动态调合方法
CN113406274A (zh) * 2021-05-28 2021-09-17 中国人民解放军空军研究院航空兵研究所 基于粘度在线检测的混合均匀度分析方法及混合装置

Also Published As

Publication number Publication date
CN109765788B (zh) 2022-03-04
JP2022528657A (ja) 2022-06-15
CN109765788A (zh) 2019-05-17

Similar Documents

Publication Publication Date Title
WO2020199665A1 (zh) 一种多目标原油调合在线优化方法
Ju et al. Extension of VIKOR method for multi-criteria group decision making problem with linguistic information
CN105740467B (zh) 一种C-Mn钢工业大数据的挖掘方法
Gong et al. Bijective soft set decision system based parameters reduction under fuzzy environments
CN104680012A (zh) 一种烧结配料计算模型
CN114971227A (zh) 一种基于marcos方法的配电网设备风险评估方法
CN112818525A (zh) 一种水电机组状态评价方法及其系统
CN106094751B (zh) 一种原料的调度方法及装置
CN116882776A (zh) 钢铁生产节能策略设置方法、装置、电子设备及存储介质
Saleh et al. Implementation of equal width interval discretization on smarter method for selecting computer laboratory assistant
He et al. Scaled prioritized geometric aggregation operators and their applications to decision making
CN104008429B (zh) 一种航空发动机装配任务优先度排序方法
CN109765789B (zh) 考虑罐底油属性的原油调合双调合头协调优化方法
CN116307381A (zh) 一种基于多属性决策的绿色产品设计方案评价方法及系统
CN114004530B (zh) 基于排序支撑向量机的企业电力信用分建模方法及系统
CN115018243A (zh) 一种混合不确定的智能工厂物料需求计划批量决策方法
Li et al. Grey-incidence clustering decision-making method with three-parameter interval grey number based on regret theory
Ran Influence of government subsidy on high-tech enterprise investment based on artificial intelligence and fuzzy neural network
Wang et al. On-line optimization model design of gasoline blending system under parametric uncertainty
Wu et al. Credit risk evaluation by improved MCDM models
CN115526057A (zh) 产品方案评价方法、系统、存储介质及设备
Hu et al. Research on Teaching Evaluation Model Based on Weighted Naive Bayes
CN112199517B (zh) 一种it/ot驱动的技术管理创新方法
Chen et al. Alternative Selection of Sea Platform Considering Multi-perspective Indices
CN116843050A (zh) 一种用于铁路货运量的预测方法、系统、设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19923595

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2021557699

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19923595

Country of ref document: EP

Kind code of ref document: A1