INTERNATIONAL JOURNAL OF ENERGY RESEARCH Int. J. Energy Res., 22, 373—381 (1998)

ON THE USE OF ECLIPSE CODE FOR OPTIMIZING INDUSTRIAL PROCESSES M. CALDAS-, J. SANTOS‡ AND V. SEMIA3 O* Mechanical Engineering Department, Instituto Superior Te´ cnico, Av. Rovisco Pais, 1096 Lisboa Codex, Portugal

SUMMARY The sequential code ECLIPSE is used in the present work to perform a technical analysis of two industrial processes—a coke gas cleansing plant and a power plant—aiming their energetic and environmental optimization. The code is validated herein comparing its results against existing experimental data acquired at the above-referred plants, for the present operating conditions. Agreement is observed to be rather good. In order to optimize both processes as far as energy and environment aspects are concerned, alternative unit operations are suggested and are included in the production flow sheet and entirely new processes are simulated. The improvements attained in both processes are noticeable. Therefore, ECLIPSE proved to be an adequate tool for global industrial processes simulation, analysis and optimization. However, the code exhibits some limitations in simulating detailed complex physical phenomena, such as combustion. ( 1998 John Wiley & Sons, Ltd. KEY WORDS ECLIPSE code; energy optimization; emissions reduction; industrial process design

1. INTRODUCTION During the last two decades, several researchers and industrial engineers have focused their attention towards the problem of energy consumption aiming its reduction, without sacrificing production, product quality and environment. In order to reduce the energetic consumption of any industrial process it is first necessary to define the process dependence on its controlling parameters, a task that is not simple. Indeed, until a few years ago, this was done by empirical methods, extrapolating data from similar equipments. It is now recognized that this approach can lead to erroneous conclusions and a new approach is required. Preferably, there should be a mathematical model able to quantify the effect of the various parameters on the global performance of the system. This approach was unimaginable until the development of high-speed digital computers due to the complexity of the phenomena occurring in general industrial processes. With the event of such computers it is presently possible to simulate an industrial process and to obtain quantitative and qualitative indicators of its performance. After completion of the actual process simulation, it constitutes a simple matter to numerically experiment changes in the operating parameters of the process aiming possible improvements.

* Correspondence to: V. Semia8 o, Assistant Professor, Mechanical Engineering Department, Instituto Superior Te´cnico, Av. Rovisco Pais, 1096 Lisboa Codex, Portugal. -Research Assistant. ‡Mechanical Engineer.

CCC 0363—907X/98/040373—09$17.50 ( 1998 John Wiley & Sons, Ltd.

Received 29 July 1997 Accepted 6 October 1997

374

M. CALDAS, J. SANTOS AND V. SEMIA3 O

In order to follow this approach several commercial computer codes have recently appeared in the market. For instance, Sundberg and Wene (1994) developed a nonlinear model of analysis and optimization of flow systems. It was named MIMES and was successfully applied in the energetic optimization of a paper mill. Bridgwater and Double (1994a) developed the AMBLE program to technically and economically simulate a vast range of technologies to obtain liquid fuels from biomass. The same authors (Bridgwater and Double, 1994b) modelled several indirect coal liquefaction processes through computational sequential codes. Goldthorpe et al. (1994) presented the ACCESS code that permits the execution of parametric studies on the thermodynamic performance of direct coal liquefaction processes, as well as economic analysis. ACCESS was successfully tested in the initial stages of the coal liquefaction systems of the British Coal Corporation. Finally, Williams and McMullan (1994) reported the development of the package ECLIPSE. This package, which runs in an ordinary personal computer under the MS-DOS environment, allows for technical and economic simulations of industrial processes, and has been tested in the coal liquefaction systems of the British Coal Corporation, providing results in good agreement with those of ACCESS. In the present paper attention is focused on the ECLIPSE code. This code is utilized to perform an energetic analysis of two industrial processes and is evaluated as a tool for energetic analysis. It should be noted that such kind of sequential codes produce results based on global mass and energy balances and are, therefore, not appropriate for detailed design of equipments. Nevertheless, in an industrial process, the replacement of a unit operation by another more efficient requires a previous study and analysis of the energetic, environmental and economic impact on the process caused by such substitution. It is for this kind of study that sequential codes constitute a powerful tool.

2. THE ECLIPSE CODE In order to perform an industrial process simulation with the ECLIPSE code it is first necessary to set the process boundaries and to establish the flow sheet. This is done in terms of modules—chemical engineering unit operations and reactors—connected together by streams: process flows. These streams are composed by a defined number of compounds named the chemical components. These compounds must be previously defined in a compound database (see ECLIPSE, 1992). It should be noted that this database does not support radicals or ions, making difficult the simulation of detailed chemical reactions, such as combustion. After the definition of the process flow diagram and the specification of the required technical data for each module, the program checks for consistency between the process technical data and the compound database. If no inconsistencies are found the equilibrium mass and energy balances are evaluated. For that, ECLIPSE has incorporated specific subroutines to calculate the necessary thermodynamic properties from the more fundamental data available at the compound database. When a convergent solution is attained the stream densities are calculated. It is then possible to calculate the process utilities requirements. The program matches the available utilities, as defined in the utilities database, to the requirements of the process and evaluates the utilities and imported fuel usages. This concludes the technical evaluation of the process. The conclusion of the technical evaluation allows for the economic calculations. To perform the economic analysis it is necessary to estimate the process and utilities capital cost. This estimation is based on the data obtained by the mass and energy balance, utilities usage calculations and on the factors and indices specified on the cost database. Additional engineering cost data is also required which varies with module type and equipment type. It is then necessary to calculate the process operating costs. These include the stream and fuel costs. As in the case of capital costs this evaluation is based on the data obtained by the mass and energy balances and utilities usage calculations, and on the factors and indices specified in the cost database. These features make of ECLIPSE an appropriate tool to analyse a process from a global point of view. However, the code is hindered to perform detailed simulations of some particular sub-processes, due to the way used by ECLIPSE to define a process in terms of modules. Each module represents a unit operation and is usually restrained by some approximations, not always reliable to the case under study. For example, each Int. J. Energy Res., 22, 373—381 (1998)

( 1998 John Wiley & Sons, Ltd.

375

ECLIPSE CODE FOR INDUSTRIAL PROCESS

chemical reaction module is either isothermal or adiabatic and only one equilibrium reaction is allowed to be defined within it. This makes it extremely difficult to perform a rigorous treatment of combustion systems without resorting to a considerable number of modules. Moreover, the sequential and iterative nature of this code can make the optimization of a process somehow lengthy, sometimes being necessary to resort to a trial-and-error approach. Due to some lack of flexibility of the input data allowed by the program, it becomes sometimes necessary to produce several outputs and adjust the input data several times. For example, if it is supposed to define a stream with given temperature and quality, this is impossible to be directly done since temperatures and pressures are the only assignable thermodynamic variables. Therefore, it is necessary to establish a routine procedure: estimation of the pressure of the stream, production of an output, check of the value of the stream quality and correction of the value of the pressure. This process has to be repeated until the quality output is the required. On the other hand, the modular and sequential structure of the program becomes quite convenient to thoroughly simulate large processes. The possibility of calculating the utilities usage with minimum effort after the achievement of a convergent solution for the mass and energy balances becomes extremely convenient. Moreover, the integration of the economical and the technical calculations provides more reliable results than those that would be produced with independent calculations.

3. THE INDUSTRIAL PROCESSES STUDIED This section presents the selected cases utilized to validate ECLIPSE code — a coke gas cleansing plant and a power plant. For both cases only energetic and environmental analysis were performed. Besides the validation of ECLIPSE by means of comparison of the predicted values against existing experimental data from the plants, the code is also used to optimize the processes, as far as energy consumption and pollutant emissions are concerned. 3.1. Coke gas cleansing plant One of the processes analysed in this work is that of a coke gas cleansing plant of the Portuguese Steel Industry (Siderurgia Nacional, Empresa de Serviios S.A.). Coke gas is a by-product of the coke production process from coal distillation. In the coke production coal is heated up to a temperature of about 1300°C in an inert atmosphere to prevent combustion. This causes the release of all its volatile components, which are collected in an adequate equipment. These volatile components are the constituents of raw coke gas. Coke gas is to be used as a gaseous fuel but, due to its high content of impurities, has to be previously cleansed. This fuel will be primarily used to heat up the coke distillation ovens and, if in excess, it will be used elsewhere in the steel plant (Figure 1). In the present application it is intended to optimize the energy consumption of the plant, keeping the gas quality standards, established on the basis of its impurity contents at the process outlet. A serious problem existing in this kind of processes is the fouling of valves and burners due to the presence of naphthalene in the

Figure 1. Simplified diagram of the coke plant under consideration ( 1998 John Wiley & Sons, Ltd.

Int. J. Energy Res., 22, 373—381 (1998)

376

M. CALDAS, J. SANTOS AND V. SEMIA3 O

Figure 2. Coke gas cleansing plant flow sheet

gas composition. It has been established that to prevent this from occurring a well-defined temperature profile must be observed, which will cause the majority of the naphthalene to condense in the appropriate locations inside the cleansing system. In order to accomplish the above-mentioned objective the layout of the system was analysed and the corresponding flow sheet was drawn. A simplified version of this flow sheet is presented in Figure 2. In the Gas Collector the raw coke gas is cooled by water injection down to a temperature low enough to prevent damaging of subsequent equipments. In the Electrostatic Precipitator dust and tar particles in suspension are removed. Further downstream, in the Naphthalene Washer, a creosote shower, creosote being a solvent for naphthalene, removes some of the remaining naphthalene. The mixture of water and ammonium, resulting from the washing operations and referred to as ammonium water hereafter, is striped in a Distillation Column Int. J. Energy Res., 22, 373—381 (1998)

( 1998 John Wiley & Sons, Ltd.

377

ECLIPSE CODE FOR INDUSTRIAL PROCESS

Table 1. Maximum relative errors between measurements and predictions Maximum mass flow error (%) 25)0

Maximum temperature error (%)

Maximum pressure error (%)

8)9

1)2

and the resulting ammonia is burned in the Ammonium Oven. All the tar removed from the gas is stored in the tar tanks and is to be sold as fuel. In order to characterize and quantify the energy consumption of the process, data referring to the month of May 1996 was collected. With this data, and after a statistical treatment, the process was analysed and quantified with resort to non-automatic calculations, in order to allow the validation of the ECLIPSE code. It should be mentioned that during the period under analysis the referred required temperature profile was not observed. Therefore, the simulation with the actual temperature profile was also performed. The agreement observed between the measurements and the predicted values is fairly good as it can be seen from Table 1. The errors presented in Table 1 were calculated according to equation (1), where measurements refer to measured and non-automatically calculated values. error "

DECLIPSE predictions!measurementsD ]100 measurements

(1)

Both pressure and temperature errors are acceptable given the uncertainties of measurements and the simplifying assumptions used in the non-automatic calculations. However, the error associated with the mass flow is larger. On one hand, this is due to the number representation used by ECLIPSE, which is of the form xxxx,xxx. This means that the smallest possible mass flow representable by ECLIPSE is limited to 1 g s~1. In the studied process there are small fluxes, the smallest being around 5 g s~1, which suffer severe rounding errors. On the other hand, the referred errors have also origin in the simplifying assumption, used in the non-automatic calculations, of disregarding the interaction between water vapour and other gas constituents. Not taking into account those two error sources, the maximum value for the error associated with the mass flow is 6)3%, which makes the prediction acceptable. The following step of the coke gas cleansing plant analysis was the simulation of the process obeying the correct required temperature profile. To achieve this goal the actual process had to be slightly modified, but the feed flows were kept unaltered. This simulation allowed for the identification of the major shortcomings of the process, both as far as energetic consumption and environmental impact are concerned. The energetic consumptions of the actual and optimized processes, both obeying the required temperature profile, are shown in Tables 2 and 3. Although the input raw coke gas is the same for both cases, the net production of cleansed gas is higher for the optimized process. This explains the differences in the specific consumptions for the same equipments for both simulations, as they are reported to a unit of cleansed coke gas. As it can be seen from Table 2 the major energy consumer is the Distillation Column, representing 76)6% of the total consumption. It should be noted that this unit operation exists only for environmental purposes, since the resulting ammonium is burned further downstream in the Ammonium Oven, which consumes cleansed coke gas. Moreover, the distillated water is not pure, but contains tar and ammonium and is discharged into the river. These are the reasons why the Distillation Column requires a detailed analysis, including the possibility of its substitution in the process, in order to make it more environmental acceptable and less energy consumer, i.e. more efficient. Another environmental flaw of the process is NO emissions x emerging from the Ammonium Oven, a point to be also dealt with in this work. ( 1998 John Wiley & Sons, Ltd.

Int. J. Energy Res., 22, 373—381 (1998)

M. CALDAS, J. SANTOS AND V. SEMIA3 O

378

Table 2. Specific energetic consumption (sc) of the original process, with the required temperature profile Equipment Fan Electrostatic Precipitator Ammonia Washer 1 Ammonia Washer 2 Naphthalene Washer

sc (kJ kg~1) 49)03

Equipment

10)34

Small Tar Tank Medium Tar Tank Tar Tanks

9)31

Mixed Tanks

10)34

Distillation Column

2)12

sc (kJ kg~1)

Equipment

sc (kJ kg~1)

+0

Pumps

188)74

5)52 13)10

Pit water

3)27

0)76

Sea water

38)80

932)41

Total sc (kJ kg~1) 1263)76

Table 3. Specific energetic consumption (sc) of the modified process, with the required temperature profile Equipment Fan Electrostatic Precipitator Ammonia Washer 1 Ammonia Washer 2 Naphthalene Washer

sc (kJ kg~1) 36)34

Equipment

sc (kJ kg~1)

Equipment

sc (kJ kg~1)

10)40

Small Tar Tank Medium Tar Tank Tar Tanks

9)36

Mixed Tanks

0)76

Pit water

0)14

10)40

Ultra filtration

9)02

Sea water

39)66

2)14

Reverse Osmosis

43)83

13)18

Pumps

185)77

+0 5)55

Total sc (kJ kg~1) 366)54

Membrane separation process appears to be the best alternative to the distillation operation. However, it is not an easy task to separate NH from H O through a membrane process, since the two molecules are very 3 2 similar: as their molecular weights, 17 and 18 respectively, can infer it. Therefore, the finest membrane process (reverse osmosis) will be required (Marr and Koncar, 1993). Ho and Sirkar (1992) showed that separation efficiencies of 90% can be attained through reverse osmosis using pressures up to 105 bar. For the present process higher efficiencies are requested, therefore it is necessary to resort to two reverse osmosis separation units installed in series, allowing the accomplishment of a separation efficiency of 99%, which is satisfactory for the present application. Reverse osmosis membranes are very sensitive to fouling and their cleansing procedure is difficult and expensive. Particularly, in the present case, the traces of tar in the ammonium water would suffice to require frequent cleansing operations. A possible solution for this problem is the re-motion of the tar from the ammonium water prior to its passage through the reverse osmosis membranes. This can be done through an ultrafiltration membrane process, which is coarser than reverse osmosis but appropriate for the purpose. Int. J. Energy Res., 22, 373—381 (1998)

( 1998 John Wiley & Sons, Ltd.

ECLIPSE CODE FOR INDUSTRIAL PROCESS

379

Table 4. Pollutant emissions in g kg~1 of cleansed coke gas for both the actual and the optimized processes Process Emissions to the River Emissions to the Atmosphere

Tar C H 2 6 NH 3 NO SO 2

Original Modified 1)241 0)069 1)172

0 0 0)069

39)655 1.448

0)416 1)387

Ultrafiltration process retains particles bigger than 10—200 As and has already been used to separate emulsions of water and machining oils (Ho and Sirkar, 1992). According to these authors, efficiencies of 85% can be achieved with operating pressures of 20 bar. Therefore, and aiming the protection of the reverse osmosis process, it is necessary to place two systems again in series, so that a separation efficiency of 97)75% is attainable. The water resulting from those processes will be almost pure water and can, therefore, be re-cycled and re-used in the ammonium washers, avoiding the need for such large amounts of fresh water in the process, as the actual plant requires. The tar removed in the ultrafiltration process can be sent to the Medium Tar Tank, resulting in a greater tar production. The concentrate resulting from the reverse osmosis process is unattractive to recover since it consists of an aqueous solution of several gases, which is very difficult to separate into its constituents. So, as before, it will continue to be burnt in the Ammonium Oven. The last problem in the coke gas cleansing process refers to the pollutants emissions from the Ammonium Oven. This oven usually operates at about 800°C. At this temperature the only significant reaction involving ammonium is its oxidation, producing NO. According to Miller et al. (1981), in the temperature range between 1000°C and 1500°C another reaction becomes important, in which NH is combined with NO 3 removing it. It seems environmentally profitable to operate the Ammonium Oven at temperatures above 1000°C. For this, the excess air coefficient was adjusted to ensure a temperature of 1250°C inside the oven. A process including all the above-proposed modifications was simulated using ECLIPSE, and the results are presented in Tables 3 and 4. As it can be seen from Table 4 the pollutants emissions are substantially reduced. 3.2. Power plant The other process analysed in this work is a power plant also belonging to Siderurgia Nacional, Empresa de Serviios S. A. This plant (see Figure 3) works as a cogeneration installation. Three different types of fuel, namely blast furnace gas, coke gas and fuel oil, can be simultaneously burnt in the boilers in order to produce steam. The first two fuels are by-products of the steel process and the third is imported. Most of the produced steam is then expanded in a steam turbine coupled to a Turbo-Alternator to produce electricity. Part of the remaining steam is expanded in another steam turbine, engaged to a Turbo-Blower group. The net work produced by this turbine will turn a compressor to produce blast to the blast furnace. The steel and ancillary processes require considerable amounts of low/medium pressure steam (4 and 15 bar), partially provided by an extraction from the turbine of the Turbo-Alternator. Expanding and cooling superheated steam from 65 bar and 450°C to the desired pressure and temperature fulfills the remaining needs. Energetically, this is obviously an inefficient process as energy is being utterly wasted. A process converting this wasted energy into a form of useful energy would be appropriated to overcome this process shortcoming. A good alternative would be the use of a back pressure steam turbine (65 bar/5 bar) with an intermediate ( 1998 John Wiley & Sons, Ltd.

Int. J. Energy Res., 22, 373—381 (1998)

380

M. CALDAS, J. SANTOS AND V. SEMIA3 O

Figure 3. Power plant flow sheet.

pressure extraction. Before the simulation of the optimized process, with the back pressure turbine, the actual process was simulated for validation and comparison purposes. The results are shown in Table 5 and compare well against experimental data obtained in the plant. It should be mentioned that for this simulation the mass flow error is zero. The reason is that all the mass flows were prescribed as an input for the simulation. The optimized process was simulated using the ECLIPSE code as before and the predicted results are presented in Table 6. As this is a cogeneration installation, the chosen parameters for comparison between alternative processes are EUF (Energy Utilisation Factor) and FESR (Fuel Energy Saving Ratio) defined by equations (2) and (3), respectively: ¼ #Q 1 EUF" / Q 4

(2)

Q ¼ 1 # / !Q 4 g g 5 FESR" " ¼ Q 1# / g g 5 "

(3)

In these equations ¼ represents the net work produced, Q is the heat required by the process, Q is the heat / 1 4 supplied, g is a typical value of a conventional boiler efficiency (0)15) and g is a typical value of a thermal " 5 cycle efficiency for energy production (0)4). Int. J. Energy Res., 22, 373—381 (1998)

( 1998 John Wiley & Sons, Ltd.

381

ECLIPSE CODE FOR INDUSTRIAL PROCESS

Table 5. Maximum relative errors between measurements and predictions Maximum mass flow error (%)

Maximum temperature error (%)

0)0

Maximum pressure error (%)

6)0

5)7

Table 6. EUF and FESR for both the actual and the optimized processes Process EUF FESR

Without back pressure steam turbine

With back pressure steam turbine

0)302 0)139

0)307 0)143

As can be seen, the EUF has increased approximately 2% with the inclusion of a back pressure steam turbine in the process, due to the increase of net work produced, which constitutes an energetic advantage. Confirming this result, the value of the FESR has also increased by 3% approximately.

4. CONCLUSIONS The ECLIPSE code has proved to be a suitable tool for global industrial process simulation. In spite of the difficulties arising in detailed simulation of equipments such as combustors, the agreement between the predicted and the experimental values is rather satisfactory, suggesting that a thorough and detailed simulation of all the physical processes involved in the process is not essential for the present kind of purposes. The improvements suggested to the studied processes were evaluated, based on predictions obtained, making resort to ECLIPSE simulations. Without using ECLIPSE, the quantification or even a realistic estimate of the benefits resulting from the alternatives presented herein would be very difficult to achieve. Although some features of ECLIPSE were not tested, namely the economic analysis and the maintenance program, since the aim of the study was purely technical, the results have shown that energetic and environmental benefits are important and technically achievable. Obviously, an economical analysis would complement the scenario and would provide a more funded decision. Nevertheless, ECLIPSE code is undoubtedly a powerful and useful tool to perform the optimization of industrial processes. REFERENCES Bridgwater, A. V. and Double, J. M. (1994a). ‘Production costs of liquid fuels from biomass’, Int. J. Energy Res., 18, 79—95. Bridgwater, A. V. and Anders, M. (1994b). ‘Production costs of liquid fuels by indirect coal liquefaction’, Int. J. Energy Res., 18, 97—108. ECLIPSE (1992). Process Simulator Manual—Version 2.1, Energy Research Centre, University of Ulster, Colerain, U.K. Goldthorpe, S. H., Cross, P. J. I. and Topper, J. M. (1994). Development of a computer package for the calculation of direct coal liquefaction process economics. Int. J. Energy Res., 18, 109—115. Ho, W. S. W. and Sirkar, K. K. (1992). Membrane Handbook, Van Nostrand Reinhold, New York. Marr, R. and Koncar, M. (1993). Recovery of ammonia from industrial wastewater, Int. J. Chem. Eng., 33, 416—425. Miller, J. A., Branch, M. C. and Kee, R. J. (1981). A chemical kinictic model for the selective reduction of nitric oxide by ammonia, Combustion and Flame, 43, 81—98. Sundberg, J. and Wene, C. O. (1994). Integrated modeling of material flows and energy systems (MIMES), Int. J. Energy Res., 18, 359—381. Williams, B. C. and McMullan (1994). Development of computer models for the simulation of coal liquefaction processes. Int. J. Energy Res., 18, 117—122. ( 1998 John Wiley & Sons, Ltd.

Int. J. Energy Res., 22, 373—381 (1998)