Send Orders of Reprints at Reprints@benthamscience.ae Dynamic Model Parameter Identification and Simulation of Scr Based on Genetic Algorithm §

The Selective Catalytic Reduce (SCR) is studied. The unknown parameters of the SCR kinetic model equations are fitted based on the Genetic Algorithm (GA), which is in the range of the allowable error, compared to the experimental data. Then in AVL Boost software, the simulation results of SCR reaction are obtained. Compared to the test data, the simulation results prove that the parameter identification is effective. At last, the SCR reaction is simulated in AVL Boost, and at the same exhaust temperature, the effect of GHSV and NSR on the SCR reaction is studied.


INTRODUCTION
Diesel engine is widely used because of its excellent dynamic performance and fuel economy, and CO, HC emission is better than that from gasoline engine. But NO X and PM emission from diesel engine is a problem, thus it is a main research direction in the field of engine to reduce NOX and PM emission from diesel engine effectively [1]. SCR as an effective method to reduce NO X emissions has been widely applied in foreign countries. Scholars from home and abroad have researched on reaction mechanism of SCR, reaction catalyst, SCR system structure and so on by means of experiments and simulations [2][3][4][5][6][7][8][9]. Simulations have aroused the attention of researchers due to its ability to reduce the test cost and accelerate research process. One of the key factors is to determine the kinetic parameters of simulation of SCR reaction, which will affect the accuracy of simulation results directly. At present, some researchers have calculated the kinetic parameters of SCR with different methods [10,11]. But these methods are based on the SCR catalyst preparation test, and this test needs more basic SCR catalyst parameters which are difficult to be obtained. In this paper SCR reaction is simulated with AVL Boost software, and the dynamic parameters of the chemical reaction are identified using the genetic algorithm, because of its the rapid convergence. After the results are calibrated, the effect of impact factors on the SCR reaction is studied.
*Address correspondence to this author at the College of Automotive Studies, Tongji University, Shanghai, 201620, P.R. China; E-mail：loudiming@tongji.edu.cn § Foundation Item: China 863 fund "Research on the key technology of diesel engine after-treatment to meet Euro emission standard VI" (2012AA 111717)

SCR Chemical Reaction
Diesel engine SCR reaction principle is that at a certain temperature and under the action of catalyst, the emissions NO X from diesel combustion take place the selective oxidation reduction reaction with reducing substance NH 3 or urea, which generates H 2 O and N 2 .
Oxidation chemical reaction of typical Urea-SCR system includes urea pyrolysis and hydrolysis, NO X and NH 3 reduction reaction, NH 3 oxidation reaction. Among them NO X and NH 3 reaction includes four types: fast SCR reaction (1), standard SCR reaction (2), Slow NO SCR reaction (3) and Slow NO 2 SCR reaction (4).
When the concentration of oxygen is over 2% in the reaction, the reaction progress is standard SCR reaction and fast SCR reaction, and in diesel engine exhaust NOx mainly exists in the form of NO, so this research mainly consider standard SCR.

SCR Reaction Model
At present there are a lot of research on the mechanism of SCR reaction, and the catalyst of SCR is V 2 O 5 that is carried on TiO 2 [12][13][14][15]. Comparing these mechanisms, the public have accepted the mechanism proposed by TopsØe. The mechanism considers that -NH 4 + is formed with NH 3 by being adsorbed on Brönsted acid sites (also known as adsorption sites) V 5+ -OH firstly, then it is oxidized to NH 3 + by adjacent oxidation potential V 5+ =O, while V 5+ =O is restored to the tetravalent state V 4+ -OH. The combination of -NH 3 + and NO from gas phase format -NH 3 + NO which is decomposed into N 2 and H 2 O instantly. V 4+ -OH is decomposed into V 3+ , V 5+ =O and H 2 O through the same responses among species, V 3+ is oxidized to form V 5+ =O under the function of O 2 , thus the catalytic cycle process is completed [12].
From this analysis, the SCR reaction is accordance with the Eley-Rideal mechanism, the standard reaction rate equation in SCR [16]: (5) is the rate of adsorption of NH 3 [16] (6) Combined with the Arrhenius equation [17], the reaction rate equation can be expressed as： Among them, r NO : the reaction rate of NO, kmol  s -1  m -3 , k 1 : the pre-exponential factor of NO reaction，s -1 , E 1 : the activation energy of the reaction of NO, J  mol -1 , C 1 : the molar concentration of NO in the reactor, kmol  m -3 , k 2 : the pre-exponential factor of the NH 3 adsorption reaction, m 3 kmol -1 , E 2 : the activation energy of adsorption reaction of NH 3 , J  mol -1 , C 2 : the molar concentration of NH 3 in the reactor, kmol  m -3 , T : the reaction temperature, K, R: ideal gas constant, 8.314J  mol -1  K -1 .

Basic Principle of Genetic Algorithm (GA)
In the natural evolution, each species become to be more adaptable to the environment. The basic characteristics of the individual species is inherited by their offspring, but they are not entirely same with their parents. The nature of the individual is decided by the chromosome which is composed by ordered arrangement gene. Individuals decided by chromosome have different adaptability to the environment, through gene hybridization and gene mutation produce the individual of strong adaptability. In the development of generation, the natural selection "the survival of the fittest" has the forces to make the individual feature be adapted for the environment.
GA is the search algorithm which is based on natural selection and natural genetic mechanism, the search space of optimization problems is mapped for the genetic space, each solution is coded as a binary string (also have other coding methods) chromosomes. Each chromosome position is called genes. Each chromosome (corresponding to an individual) represents a solution and a certain number of individuals forming a group. GA first randomly generate initial population consisted of some individuals (a group of candidate), the fitness function on the problem of environmental adaptation is determined according to the objective function. According to the fitness function of the objective function the fitness of each individual to the problem of environment is computed, then the corresponding individual chromosomes is chosen according to the individual fitness. At last genetic operation such as crossover and mutation has evolved into the generation of groups. So the generation continuously evolves towards a more optimal solution, and finally some convergence condition of adapting environment is met, and the optimal solution of the problem is obtained.

SCR Test Results
This paper takes a heavy duty diesel engine SCR as a research object, the dynamics parameters of SCR in equation (1) k 1 , E 1 , k 2 , E 2 are solved. When the four kinetic parameters are solved, the test results of other parameters in the SCR are shown in Table 1. The reaction temperature of these tests is 300℃-350℃，so the applicable temperature scope of the kinetic parameters is 300℃-350℃。

Dynamic Parameter Identification
The SCR dynamics parameter identification process based on genetic algorithm is shown in Fig. (1).
The genetic algorithm parameters are set in the Table 2.
The calculation results obtained by genetic algorithm are taken into the standard SCR kinetic equation, and the NO reaction rate is obtained. Compared to the experimental data, the effect of fitting error is within 5%, which can be accepted.
Kinetic parameters of SCR based on genetic algorithm are shown in Table 3.

The Simulation and Calibration of SCR
AVL Boost is used for the simulation of automobile engine. SCR could be simulated by the software, but the process of simulation need to input dynamic parameters. The kinetic parameters obtained by genetic algorithm are plugged into the SCR simulation model built in AVL Boost. The simulation results of SCR model are compared to the experimental results to verify the correct of reaction kinetic parameters obtained with genetic algorithm. The comparison result is shown in Fig. (2). The data issued by simulation are in line with the experimental data. The error was less than 5%. The result shows that the kinetic model of SCR reaction obtained by genetic algorithm is closed to the actual chemical reaction process, so these kinetic parameters can be used for further study.

SIMULATION OF SCR INFLUENCE FACTORS
The obtained parameters are applied into kinetic model in AVL BOOST software. The variation curve of NO conversion rate is obtained by adjusting GHSV or NSR, When the reaction temperature is in the range of 300℃to 350℃, which provides a theoretical basis for studying the influence factors of SCR reaction and improving the conversion rate of NO. Fig. (1). Flow-chart of identification process of standard scr model.   Fig. (2). NO conversion rate of experiment and simulation.

Effect of GHSV on Reaction of SCR
GHSV is defined as volume flow of exhaust gas through an unit volume catalyst under standard conditions. Reaction time is the inverse of GHSV. Keeping exhaust temperature and volume unchanged, GHSV is controlled well by adjusting exhaust gas. Fig. (3) shows the maximum of NO conversion rate when GHSV was 10000h -1 , 20000h -1 , 30000h -1 , 40000h -1 , 50000h -1 and 60000h -1 by the simulation. NO conversion rate decreases as GHSV increases, which is because of the less reaction time. The contact time between gas and catalyst become shorter, a part of gas cannot be adsorbed completely and mass transfer has been discharged, so the low conversion rate is caused. Fig. (3). NO conversion ration with GHSV.

Effect of NSR on Reaction of SCR
Keeping exhaust temperature and GHSV unchanged, NSR is controlled well by adjusting the amount of urea injected. The rate of NO conversion is shown in Fig. (4). The conversion rate of NO increases with NSR increasing. While NSR is greater than 1, the increase trend slow down, and it is close to a fixed value. This is largely because that the concentration of NH 3 increases with urea injected increasing. The conversion rate of NO improves with the ratio of NH 3 to NO increasing. While when the ration is bigger than 1, the NO conversion will be closed to limit value and no longer increases. Fig. (4). NO conversion ratio with NSR.

CONCLUSION
1) Compared to the experimental data, it is proved that it is a right way to identify the kinetic parameters of SCR with genetic algorithm. The pre-exponential factor of NO reaction is 57.22s -1 , the activity energy of NO is -0.21037J ﹒ mol -1 . The pre-exponential factor of NH 3 adsorption is 0.0027m 3 ﹒kmol -1 , the activity energy of NH 3 is 996.81J﹒mol -1 .
2) NO conversion rate is calculated with the simulation model of SCR assigned with identified parameters in AVL Boost software. The results are in well agreement with experimental results, and the error is in under 5%.
3) Through the simulation in AVL BOOST software, it is proved that the conversion rate of NO decreases significantly with GHSV increasing, and it increases gradually when NSR is raised, which provides a new way to research SCR.