Applied Energy 114 (2014) 3044

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Synthesis of heat exchanger networks featuring batch streams

Yufei Wang a, Ying Wei b, Xiao Feng a,, Khim Hoong Chu b
a b

State Key Laboratory of Heavy Oil Processing, China University of Petroleum, Beijing 102249, China Department of Chemical Engineering, Xian Jiaotong University, Xian 710049, China

h i g h l i g h t s
 Heat integration of heat exchanger networks featuring batch streams is rstly considered.  A new method based on the heat dutytime (Qt) diagram is proposed.  Energy targeting and network design can be obtained easily.  Both direct and indirect heat integration of batch streams are considered.

a r t i c l e

i n f o

a b s t r a c t
A new method based on the heat dutytime (Qt) diagram is proposed for heat integration of heat exchanger networks featuring batch streams. Using the Qt diagram method, the energy targets and the structure of the initial heat exchanger network can be easily obtained. The method can be used both for direct and indirect heat integration of batch streams. For indirect heat integration, the heat degradation of intermediate media is considered. A case study on optimizing the heat exchanger network of a hydrazine hydrate plant is used to illustrate the application of the method. The results show that integration of this heat exchanger network without considering its batch streams can limit the total energy savings. 2013 Elsevier Ltd. All rights reserved.

Article history: Received 30 May 2013 Received in revised form 27 August 2013 Accepted 17 September 2013 Available online 13 October 2013 Keywords: Heat exchanger network Graphical method Batch stream Intermediate media Energy target

1. Introduction Chemical processes can be broadly divided into continuous and batch operations. Although not common, there exist some continuous processes featuring batch streams. Notable examples include the hydrazine hydrate production process and the delayed coking process. Because chemical processes consume large amounts of nite energy resources, many heat integration techniques have been developed over the years to improve their energy efciency. For example, heat exchanger networks in numerous continuous and batch processes in the chemical industry have become highly energy efcient as a result of heat integration. Nevertheless, despite this remarkable success, heat integration analysis has not yet been applied to continuous processes featuring batch streams. Although usually only a limited number of key batch streams are present in such processes, the heat content of these batch streams could be quite substantial. As such, heat integration analysis that treats this type of hybrid processes as strictly continuous by ignoring the small number of batch streams can limit the total energy savings.
Corresponding author. Tel.: +86 15811168976.
E-mail addresses: xfeng@mail.xjtu.edu.cn, xfeng@cup.edu.cn (X. Feng). 0306-2619/$ - see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2013.09.040

Synthesis of heat exchanger networks of continuous processes has been studied extensively, either by pinch technology [1,2] or by mathematical programming techniques [3,4]. Because pinch technology offers the advantages of intuitiveness, simplicity and clarity when compared to the mathematical programming approach, it is widely used in industry. In recent development of heat exchanger networks synthesis of continuous processes, Wang et al. [5] proposed a methodology to consider heat transfer enhancement in the optimization of heat exchanger network. Zhang et al. [6] developed a method for optimizing the operation condition of heat exchanger network and distillation columns simultaneously. This methodology allows the industry to improve its economic and environment performance at the same time. Vaskan et al. [7] developed a multi-objective design method for heat exchanger network by using a MILP based model. Life cycle assessment and environment were involved in this method. Markowski et al. [8] proposed a heat exchanger network synthesis methodology considering fouling. This methodology can monitor long-term changes in the heat exchanger network efciency. With suitable adaptations, most of the heat integration methods developed for continuous processes can be used to search for heat integration opportunities in batch processes which are

Y. Wang et al. / Applied Energy 114 (2014) 3044

31

characterized by their time-dependent mode of operation. A variety of models have been developed for heat integration of batch processes since the 1980s [9], some of which are described below. (1) Time average model [10]: This model is also called pseudocontinuous process model. The heat duties of all streams in the batch process are time averaged in the production cycle, and the utility targets are then obtained by pinch technology. Because the time-dependent features of batch streams are not considered, the targets are highly ideal and can only be approached with extensive use of heat storage. (2) Time segmentation model [11]: In this model, batch streams are re-arranged in a limited manner to recover more waste heat and avoid heat storage. This method is constrained by whether the actual process allows the re-arrangement of batch streams. (3) Time and temperature cascade analysis [12,13]: The method considers simultaneously time and temperature. Heat integration in the same time interval is considered rst, followed by consideration of the time sequence. Although intermediate heat storage is included in the method, the heat degradation of intermediate media is not considered. (4) Time pinch method [14]: The heat recovery target is obtained by using time as the main constraint and heat transfer driving force as the secondary factor. The method includes direct and indirect heat recovery. The heat degradation of intermediate media is not considered. In this paper, a new graphical method based on the heat dutytime diagram will be provided for the heat integration of continuous processes featuring batch streams. The proposed method is largely based on the many heat integration concepts and tools arising from the research on continuous and batch processes. The method can be used both for direct and indirect heat integration.

nature of batch streams on the same plot. Fig. 1 shows a cold stream (Fig. 1a) and a hot stream (Fig. 1b) plotted on Qt diagrams, which use the stream heat load (Q) for the vertical axis and time for the horizontal axis. The arrowheads in Fig. 1 indicate the direction of temperature increase for the cold stream and the direction of temperature decrease for the hot stream. The temperature of the cold and hot streams increases and decreases with time, respectively. Therefore, the cold stream line has a positive slope while the hot stream line a negative slope. The vertical axis length dened by the stream boundaries gives the stream heat load while the horizontal axis length dened by the stream boundaries provides the stream time interval. Like TH diagrams, in the Qt diagrams, moving the cold and hot streams upward or downward will not affect their heat loads. They can therefore be plotted anywhere on the vertical axis. For multiple batch streams, a systematic procedure for constructing the Qt diagram is given below. (1) Calculate the heat load of each stream. (2) Rank the streams in ascending order of supply temperature. The top ranked stream is the one with the lowest supply temperature and must thus be a cold stream (If the stream with lowest supply temperature is a hot stream, it should be kept outside the heat recovery project). If two streams have the same supply temperature, rank the one with the lower target temperature rst. If two streams have identical supply temperature and target temperature, rank the one with the lower heat duty rst. (3) Plot the top ranked stream on the Qt diagram using its calculated heat duty value and time interval. Begin with the start time. Its y-coordinate (initial heat duty value) at the start time is assumed zero. Its y-coordinate (nal heat duty value) at the end time is the computed heat duty value. The plotted line will have a positive slope. (4) Plot the next stream on the Qt diagram. If it is a cold stream, begin with the start time. Its y-coordinate at the start time is given by the largest y-coordinate of the preceding stream. Its y-coordinate at the end time is given by the sum of its heat duty and the largest y-coordinate of the preceding stream. The plotted line will have a positive slope. If it is hot stream, begin with the end time. Its y-coordinate at the end time is given by the largest y-coordinate of the preceding stream. Its y-coordinate at the start time is given by the sum of its heat duty and the largest y-coordinate of the preceding stream. The plotted line will have a negative slope. Plot the remaining streams in the ranking order using the above procedure. With the four-stream example given in Table 1, let us illustrate how the Qt diagram can be constructed using the procedure described above. (1) The heat load of each stream is calculated from the following equation:

2. The heat dutytime diagram The heat dutytime diagram (Qt diagram) method developed in this work expresses the time and thermal features of batch streams intuitively and provides a practical graphical tool for heat exchanger network synthesis. As will be explained below, the method is based essentially on a combination of the Gantt chart and the temperatureenthalpy diagram (TH diagram) commonly used in traditional pinch analysis to represent continuous streams. The Qt diagram will now be illustrated by application to a batch process reported by Kemp and Deakin [12]. The stream data are given in Table 1. In this four-stream example, the batch period is 1 h with each stream only existing for a limited time period. Representing the streams graphically will allow a better appreciation of their time-dependent nature. A handy method of visualization is the Gantt chart, which is a type of time event chart. The Gantt chart is useful for visualizing which streams exist in which periods. As pointed out earlier, the TH diagram is a key tool of energybased pinch analysis which is used to represent the thermal features of continuous streams. And the proposed Qt diagram is a hybrid of the Gantt chart and the TH diagram which is able to represent the thermal features as well as the time-dependent
Table 1 Stream data. No. 1 2 3 4 Type H1 H2 C1 C2 Supply temperature (C) 170 150 20 80 Target temperature (C) 60 30 135 140

Q i CPi DT i Dti

where Qi = heat load of stream i (kW h), CPi = heat capacity ow rate of stream i (kW C1), DTi = difference of target and supply

Heat capacity ow rate (kW C1) 4 3 10 8

Start time (h) 0.25 0.3 0.5 0

End time (h) 1 0.8 0.7 0.5

Heat duty (kW h) 330 180 230 240

32

Y. Wang et al. / Applied Energy 114 (2014) 3044

Fig. 1. Qt diagrams for batch streams without phase change.

temperatures of stream i (C), and Dti = time interval of stream i (h). The computed values of heat loads are listed in the last column of Table 1. (2) Ranking the four streams in ascending order of supply temperature gives C1, C2, H2 and H1. (3) We begin by plotting the top ranked stream C1. Its start time, end time, and heat duty are 0.5 h, 0.7 h, and 230 kW h, respectively. The y-coordinate at 0.5 h is zero kW h while that at 0.7 h is 230 kW h. Hence, the coordinates of the C1 plot are given by (0.5, 0) and (0.7, 230), as shown in Fig. 4. (4) Next, we plot C2 on the Qt diagram. Its time interval is 0 0.5 h and heat duty is 240 kW h. Because C2 is a cold stream, we begin with the start time. The y-coordinate at 0 h is given by the largest y-coordinate of the preceding stream, which is 230 kW h. The y-coordinate at 0.5 h is 470 kW h, which is the sum of C1s heat duty (240 kW h) and the largest y-coordinate of the preceding stream (230 kW h). So, C2 can now be plotted in Fig. 4 using the following two points: (0, 230) and (0.5, 470). (5) H2 is the next stream to be plotted. Its time interval is 0.3 0.8 h and heat duty is 180 kW h. Because H2 is a hot stream, we begin with the end time. The y-coordinate at 0.8 h is given by the largest y-coordinate of the preceding stream, which is 470 kW h. The y-coordinate at 0.3 h is 650 kW h, which is the sum of H2s heat duty (180 kW h) and the largest y-coordinate of the preceding stream (470 kW h). Fig. 4 shows the H2 plot with the coordinates (0.3, 650) and (0.8, 470). (6) We now plot the bottom ranked stream H1. Its start time, end time and heat duty are respectively 0.25 h, 1 h, and 330 kW h. Because H1 is also a hot stream, we begin with the end time. The y-coordinate at 1 h is given by the largest y-coordinate of the preceding stream, which is 650 kW h. The y-coordinate at 0.25 h is 980 kW h, which is the sum of H1s heat duty (330 kW h) and the largest y-coordinate of the preceding stream (650 kW h). The coordinates (0.25, 980) and (1, 650) are used to plot the H1 line in Fig. 4. As can be seen in Fig. 2, the two cold steam lines have positive slopes while the two hot stream plots have negative slopes. The proposed procedure for constructing the Qt diagram gives a useful representation of when streams coexist as well as the heat duties of the streams unambiguously. To make the stream heat duties even more obvious, the actual heat duties of the streams except C1 are given in parentheses by the side of the respective heat load interval on the vertical axis (Fig. 2). However, it is not possible to plot stream temperatures on the Qt diagram. To include temperature details, the supply temperature and target temperature of

Fig. 2. Qt diagram for the four-stream example given in Table 1.

each stream are given in parentheses at the boundaries of each stream plot, as shown in Fig. 2. Note that a continuous stream at steady-state can also be plotted on the Qt diagram. In this case, the stream exists for the entire time period. It is also needed to note that in the Qt diagrams, the temperature of streams change from its supply temperature to target temperature in every time interval it exists. It does not mean the temperature of stream changes throughout from supply temperature in its start time to target temperature in its end time.

3. Application of the Qt diagram direct heat integration Heat recovery is possible when hot and cold streams of a batch process exist in the same time interval. This is known as direct heat integration. After maximum heat recovery in a time interval is achieved, the remaining heat in the hot streams is removed by cold utility and the balance of heat required by the cold streams is provided by hot utility. Heat cannot be exchanged across different time intervals, that is, a hot and a cold stream in different time intervals cannot be matched. Rescheduling of streams to allow some heat to be recovered by direct heat integration is not considered here.

3.1. Energy targets and initial network synthesis The calculation steps for direct heat integration based on the Q t diagram are as follows.

Y. Wang et al. / Applied Energy 114 (2014) 3044

33

(1) Specify DTmin and calculate energy targets: The energy targets, QHj and QCj for time interval j, are determined from the Qt diagram. For each time interval, the network is synthesized according to the principles of pinch technology (no utility coolers above the pinch, no utility heaters below the pinch, and no heat exchangers transferring heat across the pinch). P (2) Determine the energy targets for the whole cycle (QH = QHj P and QC = QCj). Direct heat integration of the four-stream example given in Table 1 will now be illustrated using the Qt diagram in Fig. 2. As can be seen in Fig. 2, there is a total of six time intervals over the batch period of 1 h. In the following calculation, a DTmin of 10 C is assumed. (1) Calculate QHj and QCj for each time interval: In this step, the time interval [0.50.7 h] is used as an example. Fig. 2 shows that three batch streams, H1, H2 and C1 coexist in this time interval. The heat load of stream i in time internal j can be calculated from the following equation:

Q ij Q i

Dt j Dt i

where Qij = heat load of stream i in time interval j (kW h), Qi = heat load of stream i (kW h), Dtj = time length of interval j (h) and Dti = time length of stream i (h). So values of Qij for H1, H2 and C1 for the period 0.50.7 h are 330(0.2/ 0.75) = 88 kW h, 180(0.2/0.5) = 72 kW h, and 230(0.2/0.2) = 230 kW h, respectively. Knowing the Qij values and temperatures of H1, H2 and C1, heat exchange matches can now be identied. Splitting C1 into three branches according to duties of the hot streams in the same time interval, one branch with a heat load of 88 kW h cools H1 to its target temperature at the end of the time interval (104 C), another with a heat load of 72 kW h cools H2 to its target temperature at the end of the time interval (54 C), and the last one with a heat load of 70 kW h is heated by hot utility. The heat capacity ow rates of the three branches are 10(88/230) = 3.83 kW C1, 10(72/ 230) = 3.13 kW C1 and 10(70/230) = 3.04 kW C1, respectively. The heat exchange matches and the heat duty of each match for this time interval are shown in Fig. 3. In this gure, numbers in bold refer to the heat duties of direct heat exchange between hot

and cold streams and the underlined number denotes the required hot utility. The direct heat integration of hot and cold streams for the other ve time intervals can be determined in the same way described above. A summary is given below. [00.25 h]: A single cold stream exists in this interval. C2 needs 120 kW h hot utility. [0.250.3 h]: One cold stream and one hot stream exist in this interval. H1 and C2 exchange 16 kW h heat (due to minimum temperature approach, they cannot exchange all 22 kW h heat), the remaining 8 kW h heat duty required by C2 is supplied by hot utility, and the remaining 6 kW h heat duty of H1 is removed by cold utility. [0.30.5 h]: One cold stream and two hot streams exist in this interval. C2 is split into two branches, one of which with a heat load of 60 kW h cools H1 and the other with a heat load of 36 kW h cools H2. The remaining heat duties of H1 (28 kW h) and H2 (36 kW h) are removed by cold utility. [0.70.8 h]: Two hot streams exist in this interval. H1 needs 44 kW h cold utility and H2 needs 36 kW h cold utility. [0.81.0 h]: A single hot stream exists in this interval. H1 needs 88 kW h cold utility. In summary, three time intervals require hot utility and four require cold utility. The results of direct heat integration for the whole cycle are shown in Fig. 4. As noted above, numbers in bold refer to the heat duties of direct heat exchange between hot and cold streams, underlined numbers signify heating utilities needed by cold streams, and numbers in normal font denote cold utilities needed by hot streams. (2) Determine the energy targets for the whole cycle (QH and QC): The hot utility target for the whole cycle, QH, can be obtained simply by summing all the underlined numbers in Fig. 4. Similarly, the minimum cold utility for the whole cycle is given by the sum of all the numbers in normal font in Fig. 4. The two overall utility targets are shown below.

QH QC

X X

Q Hj 120 8 70 198 kW h Q Cj 6 28 44 88 36 36 238 kW h

The corresponding heat recovery is 272 kW h. These targeting results are the same as those obtained by using the cascade analysis method proposed by Kemp and Deakin [12]. From the Qt diagram in Fig. 4, the structure of the initial heat exchanger network can be

Fig. 3. Direct heat integration in the [0.50.7 h] time interval.

Fig. 4. Indirect heat integration for the entire batch period [01 h].

34

Y. Wang et al. / Applied Energy 114 (2014) 3044

easily obtained, as shown in Fig. 5. Therefore, the Qt diagram approach proposed in this work is superior to the cascade analysis method. It is noted that in Fig. 5, the stream in a time interval indicates that the stream exists in this time interval, two streams will exchange heat with each other only when they both exist in that time interval. 3.2. Network optimization To reduce the capital cost of heat exchanger networks, the number of matches or units should be minimized. In general, this can be done by breaking loops and removing units. The minimum number of matches or units can be determined by Eulers network theorem. In this work, a simplied form of Eulers network theorem is used for heat exchanger network analysis, which is expressed as U = N 1. In this equation, U denotes the number of units (heat exchangers) and N indicates the number of streams including utility streams. As shown in previous sections, streams in batch processes exist in different time intervals and some streams exist across several time intervals. As a result, some of their heat exchange matches will also exist in several time intervals. For example, Fig. 4 shows that streams H1 and C2 and their matches appear in the [0.25 0.3 h] and [0.30.5 h] time intervals. When the total numbers of matches are counted by Eulers network theorem for different time intervals, those matches existing in several time intervals will be counted several times, leading to a larger number of heat exchangers. Exchangers due to miscounting are known as additional exchangers [15]. To reduce the number of matches, a useful method is to use one exchanger to exchange heat for the same hot and cold streams that exist in different time intervals, that is, additional exchangers should be removed. Accordingly, the minimum number of matches for the whole cycle can be expressed in the following way.

P network theorem for each time interval, and U+ = sum of additional exchangers. Using Eq. (3), the minimum number of matches P P for the four-stream example is Umin = jUmin,j U+ = 13 6 = 7. From Fig. 5 it can be seen that there are ve heat exchangers (E1, E2, E3, E4, E5), six coolers (E1, E3, E6, E7, E8, E9) and three heaters (E2, E4, E5) in the network, giving a total of 14 heat exchange units. This is seven units more than the minimum number of matches. It can be deduced from Fig. 5 that there are seven loops in the network, in which the four coolers for stream H1 form three, the two coolers for stream H2 form one, the two heaters for stream C2 form one, two heat exchanger pairs, E1E2 and E2E3, each forming one, and coolers E6 and E7 form one, as shown in Fig. 6. When optimizing the network, rstly, additional exchangers should be combined, and only the unit with the biggest heat load is retained. Therefore, cooler E8 is chosen for stream H1, cooler E3 for stream H2, heater E2 for stream C2 and exchanger E2 for the match between streams H1 and C2, as shown in Fig. 7. Now the network still has one loop left, that is, E8 ? E2 ? E3 ? E3. Breaking this loop will increase the utility requirements. So the loop is retained.

4. Application of the Qt diagram indirect heat integration As mentioned above, heat recovery by direct heat integration is not possible when streams do not coexist in the same time interval. To recover heat from streams that exist in different time intervals, indirect heat integration should be considered. Indirect heat integration can be realized by using thermal storage and intermediate media, that is, hot streams in a certain time interval release heat to an intermediate medium for storage, and the intermediate medium discharges the stored heat to cold streams that exist in other time intervals. In this way, heat recovery is feasible across different time intervals. Because heat transfer temperature differences are needed for the intermediate medium, compared with the original hot stream, the intermediate medium has a lower temperature, which means that using an intermediate medium will cause heat degradation.

U min U min;0

X X U min;j U
j

where Umin = minimum number of matches for the whole cycle, Umin,0 = sum of minimum number of matches counted by Eulers

Fig. 5. Initial heat exchanger network for the entire batch period.