CHEMICAL ENGINEERINGTRANSACTIONS VOL. 61, 2017 A publication of The Italian Association of Chemical Engineering Online at www.aidic.it/cet Guest Editors: Petar SVarbanov, Rongxin Su, Hon Loong Lam, Xia Liu, Jiří J Klemeš Copyright © 2017, AIDIC Servizi S.r.l. ISBN978-88-95608-51-8; ISSN 2283-9216 Relative Concentration based Mathematical Optimization for Improving Fresh Hydrogen Utilization Efficiency of Multi-Impurity Hydrogen Networks Jing Li, Qiao Zhang* School of Chemical Engineering & Technology, Xi’an Jiaotong University, Xi’an, Shaanxi Province, China, 710049 q.zhang1986@stu.xjtu.edu.cn The traditional methods of hydrogen network integration generally minimize hydrogen consumption and rarely discuss the characterization of hydrogen utilization efficiency (HUE). And some definition of HUE are also not accurate enough. Based on relative concentration property, this paper constructs a nonlinear mathematical model encompassing all necessary constraints to synthesize the multi-impurity hydrogen networks. Result shows that this model is superior to all ones with absolute concentration basis. Through observation and analysis of hydrogen to oil ratio (HTO), a practical and important operation parameter of refinery, we can know the actual hydrogen demand of one hydrogen network. So, in this paper we compare two different kinds of HUE. Conclusion is: Efficiency2, calculated by actual hydrogen demand, is more reasonable than Efficiency1, calculated by hydrogen discharge. 1. Introduction Nowadays refinery hydrogen consumption increases by about 8% a year. Therefore, hydrogen resources saving are of great significance. Alves and Towler (2002) used the diagram of surplus hydrogen to target minimum fresh hydrogen consumption. But this method computation efficiency is very low. EI-Halwagi et al. (2003) put forward a graphical method with contaminant versus flow rate diagram. Then, Foo et al. (2006) put forward algebraic calculation instead of visual solution and ultimately achieved the purpose of maximum resource recovery. Zhang et al. (2011) put forward an improved graphical method for the hydrogen networks with purification. Liu et al. (2013) further performed the research on the relationship between the Pinch Point location and the inlet, outlet concentration of purification equipment. As for mathematical programming methods, they usually build superstructure and a corresponding mathematical model. Hallale and Liu (2001) first established a superstructure model to optimize the hydrogen network. Zhou et al. (2012) added a desulfurization device. Gradually, other factors such as uncertainty (Zuwei et al. 2010), multi-objective (Jiao et al. 2013), multi-period (Liang et al. 2016), etc. were considered. With reference to water networks, Zhang et al. (2016) established relative concentration constraints and further reduced hydrogen consumption. In this paper, a multi-impurity, relative concentration basis, nonlinear programming mathematical model is established to analysis two different characterizations of HUE. According to the actual operation parameter, HTO, the best one is selected to explain and analyze the ability and potential of hydrogen resources saving. 2. Theory 2.1 Superstructure Superstructure showed in Figure 1 includes hydrogen sources (sr), sinks (sk), and purifier psa (Pressure Swing Adsorption). {1,2,...i...a} is the number of sources and {1,2...j...b} is the number of sinks. Each source can supply hydrogen to any sink and each sink can also receive hydrogen from any source. Besides, sr could part or all be sent to one or more sk, or similarly to psa or fuel gas. DOI: 10.3303/CET1761095 Please cite this article as: Li J., Zhang Q., 2017, Relative concentration based mathematical optimization for improving fresh hydrogen utilization efficiency of multi-impurity hydrogen networks, Chemical Engineering Transactions, 61, 583-588 DOI:10.3303/CET1761095 583 Figure 1: A superstructure of hydrogen network with purification The hydrogen demand of skj can be satisfied by one or more sources and/or psa. This paper builds mathematical model with this superstructure, and the target is to minimize fresh hydrogen consumption and maximize HUE. 2.2 Mathematical model 2.2.1. Calculation of relative concentration The main reference of relative concentration (RC) in this paper is research of Zhang et al. (2016). iysr i,km i,kRCsr = (1) jysk j,kn j,kRCsk = (2) mi,k and nj,k are absolute concentration (AC) of contaminant k in sri, and the ceiling AC of k in skj. RCsri,k, RCskj,k are corresponding RC. Compared with AC, RC is used to quantify the impurity concentration and the hydrogen flow rate is employed to quantify flow rate. It releases the total flow rate and concentration normalization constraints. In other words, it is relaxation, so it is superior to absolute basis. 2.2.2. Constraint equations of sr The main constraint for every source is availability of hydrogen resource. iysriFsriHsr •= (3) iHsr b 1j j,idj,iHyisiZ ≤∑ = •+• (4) The sum of hydrogen flow rate Zi (from sri to psa), Hyj,i (from sri to skj) should be less than total hydrogen flow rate of sri (Hsri). In Eq(4), si is the connection relationship between sri and psa; di,j is matrix of network structure. 2.2.3. Constraint equations of sk For every sink, there are two constraints: first is amount of hydrogen provided by sr and psa must meet the demand of skj; Second is the impurity concentration should be less than the upper limit of skj. )j+e1(j=XskjX • (5) Xskj is the current oil processing capacity, when there is a fluctuation ej in skj, it turns to Xj. jcjXjljHsk +•= (6) Hskj is the actual hydrogen demand of skj, and lj, cj are known correction coefficients. 584 ∑ ∑≤ a 1i ) a 1i j,idj,iHyjssjO(j,kRCski,kRCsrj,idj,iHykRCpsajssjO = = •+••••+•• (7) ∑ ≥ a 1i jHskj,idj,iHyjssjO = •+• (8) Oj is hydrogen flow rate from psa to skj, ssj is the connection relationship. RCpsak is RC of contaminant k in psa outlet. Eq(7) is impurity constraint of skj. Eq(8) means the constraint of hydrogen demand of skj. 2.2.4. Constraint equations of psa Constraint for psa is about mathematical relationship with its inlet and outlet. ∑∑ b 1j jssjO=1λpsa a 2i isiZ = •• = • (9) kRCpsa)jss b 1j jO(= a 2i 2λpsai,kRCsrisiZ •• == ••• ∑∑ (10) Eq(9) is psa hydrogen flow rate relationship and Eq(10) is impurity relationship both between psa inlet and outlet, where psaλ1, psaλ2 are constants of psa hydrogen recovery rate, impurities removal rate. 2.2.5. Objective function In this paper, objective function is minimum hydrogen consumption: FSRH. ∑ b 1j 1ysr j,1d1j,Hy FSRH= = • (11) And at the same time, we use following variables to do some useful analysis: jX a 1i j,idj,iHyjssjO jHTO ∑ = •+• = (12) HTOj is current hydrogen to oil ratio of skj. 1ysrFSRH )) a 2i b 1j j,idj,iHy+isi-(ZiHsr( b 1j jssjO- a 2i isiZ efuel= • = = ••+ = • = • ∑ ∑∑∑ (13) -efuel= Efficiency 11 (14) efuel is fuel gas to fresh hydrogen consumption ratio. Efficiency1 is one HUE. 1ysrFSRH a 2i iHsr- b 1j jHsk =2Efficiency • == ∑∑ (15) Efficiency2 is HUE calculated by the actual hydrogen demand. The relative concentration model consists of Eq(1) ~ Eq(15), which is a nonlinear programming (NLP) model. 3. Case studies Case in this paper is based on actual data of a refinery. Table 1 shows the detail data of sources and sinks. Table 1: Data of sr&sk in multi-impurity hydrogen network F (Nm3/h) ysr (mol%) mj,i (mol%) Xskj (t/h) ysk (mol%) nj,i (mol%) H2S N C H2S N C Hydrogen sources Hydrogen Sinks sr1 0.995 0.0009 0.003 0.0011 sr2 16000 0.915 0.007 0.016 0.062 sk1 35 0.92 0.032 0.017 0.031 sr3 90000 0.899 0.0216 0.063 4 0.016 sk2 13 0 0.91 0.012 0.0554 0.024 sr4 32000 0.88 0.037 0.041 0.042 sk3 65 0.898 0.021 0.023 0.058 sr5 7500 0.9 0.056 0.022 0.022 sk4 22 0.918 0.032 0.021 0.029 sr6 6500 0.87 0.026 0.083 0.021 sk5 18 0.883 0.022 0.065 0.03 sr7 24600 0.88 0.037 0.041 0.042 sk6 50 0.898 0.021 0.023 0.058 sr8 8864 0.87 0.026 0.083 0.021 sk7 30 0.918 0.032 0.021 0.029 sr9 14000 0.44 0.123 0.166 0.271 585 When fluctuations in sk2, sk3, sk5, use 2.2 model we get Table 2. Results of AC is got by traditional methods. There are 9 sources and sr1 is fresh hydrogen, as well as 7 sinks. And it is a multi-impurity hydrogen network. Figure 2 (a), Figure 2 (b) and Figure 3 (a) are constructed by the results in Table 2, where the X axis is different Xj, Y axis is hydrogen utilization efficiency. In Figure 2, (a) is about Efficiency1, while (b) is about Efficiency2. Table 2: Results by relative concentration basis and absolute concentration basis e2, e3, e5 -0.2 -0.16 -0.12 -0.08 -0.04 0 0.04 0.08 0.12 0.16 0.2 RC FSRH (Nm3/h) 52,993 55253. 57513 59,773 62,033 64,293 66,553 68,813 71,285 73,364 76,783 Efficiency1 0.326 0.407 0.481 0.551 0.616 0.676 0.733 0.786 0.833 0.882 0.913 Efficiency2 0.326 0.407 0.481 0.551 0.616 0.676 0.733 0.786 0.833 0.882 0.913 AC FSRH (Nm3/h) 55,650 58044 60437 62,830 65,425 68,631 71,831 76,471 80,359 82,359 85,383 Efficiency1 0.396 0.4713 0.541 0.606 0.664 0.714 0.759 0.786 0.818 0.865 0.901 Efficiency2 0.3111 0.3881 0.459 0.526 0.585 0.636 0.681 0.710 0.743 0.790 0.825 Figure 2: (a) Efficiency1 based on different Xj, Figure 2: (b) Efficiency2 based on different Xj 586 Figure 3 (a) is HUE data with the same FSRH. The main reason of why Efficiency1 is so close to Efficiency2 based RC basis is, RC basis has almost reached the limit of hydrogen savings potential. With the data of HTOj, Figure 3 (b) shows the relationship among current, optimized and minimum HTOs in blue triangle, red round and black square curves in sk3. From Table 2 we can see that compared with absolute concentration basis, the RC model consumes less fresh hydrogen and it can increase HUE (Efficiency2) by about 5%. So, RC constraint is more economical. Also in Figure 3 (a), we can see that with the same FSRH, efficiency of RC model is the highest. However, Efficiency1 is higher than Efficiency2 based on the same AC constraints. And Figure 2 (a) shows that for Efficiency1, RC basis is smaller than AC basis. Consequently, Efficiency2 is more logical. Figure 3 (b) explains the origin of fresh hydrogen conservation, that is lowering HTO within feasible region. Specifically, for the same sink, the HTO curve based on RC is below absolute concentration basis while above the lower limit. This means that absolute concentration constraint always provides more hydrogen than that of RC constraint. HTO can make us easily know that how much the fresh hydrogen savings potential is. Figure 3: (a) Different efficiency based on the same FSRH Figure 3: (b) The HTO data of sink 3 587 4. Conclusions Through the analysis of the results we summarize that the NLP model proposed in this paper is superior to all absolute concentration based methods; Efficiency2 calculated by the actual hydrogen demand is more reasonable; HTO is the primary reason to both amount and efficiency of fresh hydrogen consumption. Besides, the gap between current HTO and its lower bound represents the hydrogen saving potential of synthesis methods of hydrogen networks. Acknowledgments The financial support for this research provided by the National Science Foundation of China under Grant 21506169 and the China Postdoctoral Science Foundation under Grant 2016T90924 is gratefully acknowledged. References Alves J. J., Towler G. P., 2002, Analysis of refinery hydrogen distribution systems, Industrial & Engineering Chemistry Research, 41(23), 5759-5769. El-Halwagi M., Gabriel F., Harell D., 2003, Rigorous graphical targeting for resource conservation via material recycle/reuse networks, Industrial & Engineering Chemistry Research, 42(19), 4319-4328. Foo D. C. Y., Kazantzi V., El-Halwagi M. M., Abdul Manan Z., 2006, Surplus diagram and cascade analysis technique for targeting property-based material reuse network, Chemical Engineering Science, 61(8), 2626-2642. Zhang Q., Feng X., Liu G. L., Chu K. H., 2011, A novel graphical method for the integration of hydrogen distribution systems with purification reuse, Chemical Engineering Science, 66(4), 797-809. Liu G. L., Li H., Feng X., Deng C., Chu K. H. 2013, A conceptual method for targeting the maximum purification feed flow rate of hydrogen network, Chemical Engineering Science, 88, 33-47. Hallale N., Liu F., 2001, Refinery hydrogen management for clean fuels production, Advances in Environmental Research, 6(1), 81-98. Zhou L., Liao Z., Wang J., Jiang B., Yang Y. 2012, Hydrogen sulfide removal process embedded optimization of hydrogen network, International Journal of Hydrogen Energy, 37(23), 18163-18174. Zuwei L., Yi Z., Gang R., Yongrong Y., 2010, Improving refinery profits via fine management of hydrogen networks, China Petroleum Processing & Petrochemical Technology, 12(2). Jiao Y., Su H., Hou W., Li P., 2013, Design and optimization of flexible hydrogen systems in refineries, Industrial & Engineering Chemistry Research, 52(11), 4113-4131. Liang X., Kang L., Liu Y., 2016, The flexible design for optimization and debottlenecking of multiperiod hydrogen networks, Industrial & Engineering Chemistry Research, 55(9), 2574-2583. Zhang Q., Yang M., Liu G., Feng X., 2016, Relative concentration based pinch analysis for targeting and design of hydrogen and water networks with single contaminant. Journal of Cleaner Production, 112, 4799-4814. Zhang Q., Song H., Liu G., Feng X., 2016, Relative concentration-based mathematical optimization for the fluctuant analysis of multi-Impurity hydrogen networks, Industrial & Engineering Chemistry Research, 55(39), 10344-10354. 588