Multi-objective Optimization of Coal-fired Boiler ... - Semantic Scholar
1300
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
Multi-objective Optimization of Coal-fired Boiler Combustion Based on NSGA-II Tingfang Yu, Hongzhen Zhu School of Mechanical and Electronic Engineering, Nanchang University, Nanchang, Jiangxi Province, China; Email: [email protected], [email protected]
Chunhua Peng School of Electrical & Electronics Engineering, East China Jiaotong University, Nanchang, Jiangxi Province, China Email: [email protected]
Abstract—NOx emission characteristics and overall heat loss model for a 300MW coal-fired boiler were established by Back Propagation (BP) neural network, by which the the functional relationship between outputs (NOx emissions & overall heat loss of the boiler) and inputs (operational parameters of the boiler) of a coal-fired boiler can be predicted. A number of field test data from a full-scale operating 300MWe boiler were used to train and verify the BP model. The NOx emissions & heat loss predicted by the BP neural network model showed good agreement with the measured. Then, BP model and the non-dominated sorting genetic algorithm II (NSGA-II) were combined to gain the optimal operating parameters which lead to lower NOx emissions and overall heat loss boiler. The optimization results showed that hybrid algorithm by combining BP neural network with NSGA-II can be a good tool to solve the problem of multi-objective optimization of a coal-fired combustion, which can reduce NOx emissions and overall heat loss effectively for the coal-fired boiler. Index Terms—Coal-fired boiler combustion, Multi-objective optimization, BP neural network, NSGA-II
I. INTRODUCTION In China, the requirements for environmental protection are increasingly strict, especially for coal-fired utility boiler. Coal remains the primary energy resource in China, and one of the major concerns associated with coal-fired power plants is the emission of pollutants, especially for NO2 and NO (collectively referred to as NOx). Today, NOx emission is regulated and has become an important consideration in the design and modification of coal-fired utility boiler [1-2]. However, many olddesigned utility boilers in China emit the NOx pollutants above the limit and have posed terrible threat to the surrounding environment, coal-fired power plants face important challenges concerning the methods and technologies to meet these new environmental requirements. In addition to the developments in the plant construction and flue gas cleaners such as Selective Catalytic Reduction (SCR) reactor, NOx control techniques based on combustion modification are of considerable interest[3-5], because they avoid or postpone large capital expenditures while meeting
© 2013 ACADEMY PUBLISHER doi:10.4304/jnw.8.6.1300-1306
environmental compliance requirements compared with the relatively expensive flue gas NOx reduction technologies. The control of the boiler operating conditions through combustion optimization is an important and cost-effective way to affect NOx emissions [3-5]. In recently years, many scholars applied artificial intelligent methods to optimization of coal-fired boiler combustion, H. Zhou [6] established the boiler fly ash carbon content model to predict the relationship between unburned carbon in fly ash and operation parameters of boiler by using BP neural network , Fuzzy neuralnetwork was proposed to model NOx emissions by Ikonen [7]. P. H. Wang [8] established artificial neuralnetworks model of a boiler NOx emissions and efficiency, based on the above model, genetic algorithm (GA) was introduced to get the optimal operation parameters of a coal-fired boiler. C. XU [9] set up the boilers efficiency and NOx emissions model by minimal resource allocating networks (MRAN), simulation studies on global optimization on efficiency and low NOx emissions object were carried out based on MRAN and GA method. A sharing LSSVM model was established by F. Gao [10] to predict the NOx emission and thermal efficiency of a 1000MW boiler, and then adopted an improved PSO algorithm to optimize the operational conditions of the boiler. C. L. Wang [11] using support vector machine (SVM) combined with the ant colony algorithm to optimize the NOx emissions for a coal-fired boiler. However, coal-fired boiler is a very complex system, the level of unburned carbon in fly ash is an important factor affecting the efficiency of pulverized coal fired boiler, and especially those equipped with low NOx burners. Due to the reduced mixing intensity and the formation of fuel rich zones under low NOx combustion conditions, the residence time of the coal particles in the oxygen rich environment decreases, resulting in an increase of the amount of unburned carbon in fly ash. Pollution formation and carbon burnout in pulverized coal combustion are dominated by the fuel properties (reactivity, volatiles, nitrogen content, etc), fuel preparation (coal fineness) and combustion conditions (mixing). The emission of NOx from boiler can be
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
reduced significantly by modification of the combustion process. NOx emission from utility boilers and the boiler efficiency are the different functions of fuel properties, boiler design and operating variables. In the combustion optimization ,the conflict between low NOx emission and high boiler thermal efficiency encounters[4,5], the operation parameters suited to lower NOx emissions of coal-fired boiler always lead to a higher carbon content in fly ash and lower efficiency of the boiler, the above studies on optimization of coal-fired boiler combustion mainly focus on the single-objective optimization ,for only on the emissions of NOx from the boilers or only boiler efficiency alone, and can’t reach both lower NOx emissions and higher efficiency of boiler ,It is imperative to find a good tool for multi-objective optimization of coal-fired boiler combustion. The traditional approach is converting the multiobjective problem(MOP) into single objective problem(SOP) by means of assign different weights to all objectives , then get the optimal point by means GA or other algorithms as in the literature[8,9], its disadvantage include: a) it can’t get a set of Pareto optimal solutions but a single solution, b)different objectives is not comparable, thus a reasonable weights does not exist, the assigning of weights introduced human factors inevitably ,which may be debatable. The real result of multi-objective problem solving should be a set of Pareto optimal solutions instead of just to get only an optimal solution [12-14]. There is no single solution that is optimal (global minimum or maximum) with respect to all the optimization objectives under multi-constraints for the multi-objective optimization contrasted to the singleobjective one, and only acceptable non-dominated solutions exist, which are the so-called Pareto optimal solutions[15,16]. The approaches for solving conventional multi-objective problems include the objective weighted method, the hierarchical optimization method, the goal programming method, etc. These algorithms deal with the problems converting the multiobjective problem to the single-objective problem with the emphasis of one particular Pareto-optimal solution at a time. Actually, they are not real multi-objective optimization algorithms and cannot be widely used in solving the complex problems. Over the past decades, to solve the complicated multi-objective optimization problems and achieve the Pareto optimal solution set, a number of multi-objective evolutionary algorithms have been suggested, such as the niched Pareto genetic algorithm (NPGA)[17], the Pareto archived evolution strategy (PAES)[18], the strength Pareto evolutionary algorithm(SPEA)[19], Differential Evolution algorithm[20,21] and the non-dominated sorting genetic algorithm (NSGA)-II[12-14]. Among them, the recently well-known NSGA-II has drawn more attention due to its feasibility for any number of objectives. In this paper, in order to get the minimum NOx emissions and maximum boiler efficiency, taking the minimum the overall heat loss of the boiler and the minimum NOx emissions as the optimization objectives.
© 2013 ACADEMY PUBLISHER
1301
a BP neural-network were proposed to model NOx emissions and overall heat loss of the coal-fired boiler ,then the NSGA-II is introduced as multi-objective optimization algorithms combined with BP neural network model to obtain the Pareto optimal solution set of the multi-objective optimization of coal-fired boiler combustion problem. The rest of the paper is organized as follows: Section II describes mathematical model of NOx emissions and boiler heat loss established by BP neural network based on the test data of a 300MWe coal-fired boiler, the simulation results of BP model were given, and Analyzed the accuracy of the model. Section III introduced NSGAII algorithm and steps involved in implementation to the multi-objective optimization of a coal-fired boiler combustion problem. Section IV performed application of NSGA-II to multi-objective optimization of coal-fired boiler combustion, numerical results were given under the rated load of 300MW and a partial load of 200MW, Pareto-optimal set of multi-objective optimization on the coal-fired boiler obtained by NSGA-II were illustrated, the NOx emission and the overall heat loss of boiler shown a different variation law under the partial loads conditions and rated load, then the reason for the above laws were illustrated , finally in Section V summaries are given. II. MATHEMATICAL MODEL A. Description of the Boiler The boiler is a 300 MW, tangentially fired dry bottom boiler with double furnaces, equipped with indirect-fired coal-storage pulverizing system and exhaust pneumatic convey system, four DTM350/600 drum ball mill, four exhaust fans,32 powders, using horizontal and louver type concentrated diluted pulverized coal burner. The maximum continuous rating (MCR) of the boiler is 1025 t/h of superheated steam at 538℃. The boiler had four levels of primary air nozzles (A, B, C, D) and five levels of secondary air nozzles (AA, AB, BC, CD, DE), two levels of over-fire air nozzles(OFA1,OFA2), a level separated over-fire air nozzles(SOFA). The Cross-section of the boiler furnace and the arrangement of the boiler burner nozzles are shown as figure 1 and figure 2
Figure 1. Cross-section of the boiler furnace
1302
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
neuron in the output layer, according to a large number of simulation , we finally select 21 neurons in the hidden layer for the BP model show as in fig.3.
Predict NOx values by BP model/10-4 mg/m3
C. Simulation Results Training the BP neural network is an important step for developing a useful network. On the basis of 20 test data of the 300MWe boiler in the literature [3], the experimental data of Cases 1–18 are used as the learning samples to train the BP neural network. 2 others are employed as input to verify the accuracy of the model. The moment attachment and the self-adaptive learning step size method are adopted in the training process. The momentum factor is taken as 0.5. The training will stop when the mean square error of the system is less than 106, a total of 20000 epochs are needed to achieve the correct weight and threshold values.
Figure 2. Arrangement of the boiler burner bozzles
Velocities of primary air (4 variables) Speed of four levels powders (4 variables)
NOx
Opening of OFA1/OFA2/SOFA (3 variables) Velocities of secondary air (5 variables) Oxygen contend(1 variable)
Heat loss
9.5
Training samples Verify samples
9
8.5
8
7.5
7
6.5
6
5.5
5
BP neural network mathematical model of the coal-fired boiler combustion characteristics
B. BP Model of NOx Emissions and Boiler Heat Loss As BP neural network has good nonlinear mapping ability, the paper introduced the BP neural network to establish mathematical model of the coal-fired boiler combustion characteristics, which is used to evaluate the NOx emissions (at 6 vol%O2 dry) from the boiler and the overall heat loss of boiler. Moreover, the model can be employed as the objective functions in the later multiobjective optimization of coal-fired boiler combustion. On the basis of test data of the 300MWe boiler[3], BP neural network mathematical model of the coal-fired boiler combustion characteristics under the steady-state operation conditions was established, 17 parameters closely related to the NOx emissions and the overall heat loss from the boiler were selected as the input variables: four levels primary air average velocities(A-D), four levels average speed of the powders(A-D), five levels secondary air average velocities(AA-DE), three opening percent of the over-fire air nozzles (OFA1/OFA2/SOFA) and the oxygen content in the exhaust gas from the boiler. The output of the BP model are NOx emissions and the overall heat loss from the boiler, thus the BP neural network has 17 neurons with input layer, two output
© 2013 ACADEMY PUBLISHER
5
Figure 4.
5.5
6
6.5
7
7.5
8
8.5
9
9.5
the measured NOx values /10-4
Simulation results of the NOx emissions from the boiler by BP model
11.5
11
Predict values by BP model /%
Figure 3.
Training samples Verify samples
10.5
10
9.5
9
8.5
8
7.5 7.5
8
8.5
9
9.5
10
10.5
11
11.5
The measured overall heat loss values/% Figure 5. . Simulation results of the boiler heat loss by BP model
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
After the BP model training was completed, the simulation results of the NOx emissions from the boiler by BP model are shown in figure 4, and simulation results of the boiler heat loss by BP model are illustrated in fig.5. The figure shows that BP neural network model predict the NOx emissions from the boiler are not more than 1.8% deviation compared with the measured values, and the overall heat loss are not more than 2.3%, which can satisfied the accuracy requirement of the engineering. III. PROPOSED SCHEME A. Description of the NSGA-II The NSGA proposed by Srinivas and Deb [22] is an efficient sorting algorithm and developed from the simple genetic algorithm only in the way of the selection operator. The NSGA can simplify multi-objective optimization problems into one fitness function problem and also be used for any objective of optimization problems. The main ideas of the NSGA are as follows: First, the individuals of the population must be classified into different non-dominated sets according to the nondomination relationship. Second, to maintain the diversity in the population, sharing methods are adopted to control the individual distribution of every non-dominated set. Third, the selection operator is performed on the basis of an individual’s domination set and niche count. Finally, combined with other operators of the simple genetic algorithm, the Pareto-optimal solutions can be obtained. Deb et al. [13] carried out a substantial improvement of the NSGA and proposed a fast and elitist NSGA (i.e., NSGA-II) with the following three aspects of improvement: (i) A fast non-dominated sorting approach taken to reduce the computational complexity from O(MN3) to O(MN2), where M is the number of the objectives, and N is the population size, (ii) The elitism strategy introduced to improve the robustness and convergence, (iii) The sharing methods replaced by the crowdingdistance methods to avoid specifying the sharing parameter δ share. Combined with the tournament selection method, the simulated binary crossover (SBX) operator, the polynomial mutation operator, and the efficient constraint-handling approach, the NSGA-II displays good convergence and robustness [13]. 1) The SBX crossover operator SBX crossover operator was adopted in NSGA-II, cross-point position was received randomly, and the genetic codes of two parent individuals on both sides of the crossing point were exchanged. In general, SBX crossover puts the stress on generating offspring near the parents. This crossover guarantees that the extent of the children or offspring is proportional to the extent of the parents, and also favors that near parent individuals are monotonically more likely to be chosen as children than individuals distant from the parents in the solution space. The procedure for finding the offspring solutions from parent solutions is given in [13]. © 2013 ACADEMY PUBLISHER
1303
2) Polynomial mutation The probability of creating a solution yit+1 near to the parent is higher than the probability of creating one distant from it. The shape of the probability distribution is directly controlled by an external parameterηm and the distribution remains unchanged throughout the iterations. Like in the SBX operator, the probability distribution can also be a polynomial function, instead of a normal distribution [23]: yit+1=xit+1+(xi(U)−xi(L) )δi
(1)
where the parameterδi is calculated from the polynomial probability distribution, xit+1 and xi are one of the offspring solutions and parent solutions respectively, superscript U and L represents the upper and lower limit for the corresponding variables. P(δi)=0.5(ηm +1)(1−|δi |)ηm
δi
=(2ri)1/ηm −1 −1, if ri < 0.5 =1−[2(1−ri)]1/ηm +1, if ri ≥0.5
(2)
For handling the bounded decision variables, the mutation operator is modified for two regions: [xi(L), xi] and [xi, xi(U)] [23]. B. Algorithm Implementation Fig. 6 explains the implementation of NSGA-II algorithm for the multi-objective coal-fired boiler combustion problem. It is described in the following steps: Step 1: Identify the control variables limits of the boiler operational parameters like velocities of primary air, speed of powders, oxygen contend and velocities of secondary air for the multi-objective optimization of a coal-fired boiler combustion problem. Step 2: Select the parameters like number of population, maximum number of generation, crossover and mutation probabilities. Step 3: Generate initial population. Step 4: Evaluate objective functions (i.e. NOx and Heat loss of the boiler) by the BP model in section II for initial population. Step 5: Set the generation count. Step 6: Perform the SBX crossover operator and polynomial mutation for the set of individuals. Step 7: Perform non-dominated sorting [13] (i.e., sorting the population according to each objective function value in ascending order of magnitude). Step 8: Generate population for next generation from combined parent and offspring population. Step 9: Perform selection based on tournament selection [13] thereby a higher fitness is assigned to individuals located on a sparsely populated part of the front. Step 10: Increment the generation count and repeat the steps from 6 to 9 until the count reaches the maximum number of generation.
1304
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
6.4
NOx emissions form the boiler 10-4 mg/N3)
Identify the control variables limits of velocities of primary air, speed of powders，oxygen contend and velocities of secondary air
Select the parameters like population size, maximum number of generation, crossover and mutation probabilities
Generate initial population
Evaluate objective functions (NOx and Heat loss of boiler) for initial population
Set the generation count, i = 0
6.3 6.2 6.1 6 5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 5.1 5
Generate offspring population using SBX crossover operator and
4.9 8
polynomial mutation
8.1
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
9
9.1
The overall heat loss of boiler (%)
Perform non-dominated sorting for combined parent population and offspring population
Figure 7. Pareto-optimal solution set of multi-objective optimization of a coal-fired boiler combustion (300MW) i=i+1
Generate population for next generation from combined parent and offspring population
6.8
Perform parent selection based on tournament selection
Stop
i < i max
Figure 6. Flowchart of NSGA-II algorithm for the multi-objective coal-fired boiler combustion problem.
IV. NUMERICAL RESULTS The optimization objectives of the multi-objective optimization of coal-fired boiler combustion include the minimum of NOx emissions from boiler and the maximum of the boiler efficiency (100%-overall heat loss%), thus taking the minimum the overall heat loss of the boiler and the minimum NOx emissions as the optimization objectives in this paper.
NOx emissions form the boiler 10-4 mg/N3)
6.7 6.6 6.5 6.4 6.3 6.2 6.1 6 5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 5.1 8.6 8.7 8.8 8.9
9
9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 10 10.1 10.2
The overall heat loss of boiler (%)
Figure 8. Pareto-optimal solution set of multi-objective optimization of a coal-fired boiler combustion (200MW)
TABLE I. . OPERATION CONDITIONS OF THREE PARETO SOLUTIONS(300MW)
a b c
Primary air velocity (m/s) A B C D 28.2 28.5 28.6 29 28.6 28.5 29.1 29.3 29.2 28.9 28.9 29.5
Speed of powders(rpm) A B C D 437 402 339 335 442 395 353 334 446 392 351 322
TABLE II.
a b c
Primary air velocity (m/s) A B C D 25.5 26.1 25.8 26.2 26.7 26.5 26.6 27.1 28.1 28.9 29.5 28.8
© 2013 ACADEMY PUBLISHER
Secondary air velocity (m/s) Opening of OFA(%) Oxygen Heat loss AA AB BC CD DE OF1 OF2 SOFA % % 30.1 32.6 31.2 31.8 31.6 73 75 55 4.03 9.02 31.3 31.7 32.3 32.2 32.5 86 85 61 4.65 8.47 32.3 32.8 33.6 33.8 34.7 92 95 82 5.12 8.02
NOx 10-4 5.01 5.55 6.35
. OPERATION CONDITIONS OF THREE PARETO SOLUTIONS (200MW)
Speed of powders(rpm) A B C D 437 386 329 315 442 365 323 321 449 360 324 312
Secondary air velocity (m/s) Opening of OFA(%) Oxygen Heat loss AA AB BC CD DE OF1 OF2 SOFA % % 27.2 26.6 26.2 28.0 20.6 43 45 0 5.06 8.73 27.5 27.9 28.1 29.7 22.7 48 51 0 5.84 9.55 28.6 29.8 31.3 30.9 21.5 52 56 0 6.65 10.15
NOx 10-4 6.48 5.65 5.39
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
In the process of optimization with the NSGA-II algorithm, the premises include maintain the stability of the boiler load, the same coal injected into the boiler. Moreover, In order to ensure optimal accuracy and reasonableness of the optimization results ,the are also some constraints must be taken into account, those constraints include: a) Upper and lower limits of the optimization variables, the range of optimization variables are: primary air velocity of 25~30m/s, the secondary air in the AA~DE levels were 25~40m/s, OFA dampers opening as 0~100%, oxygen to optimize rang of 3% to 6.5%, the speed of powders ranges from 300~500rpm. b) Oxygen in the exhaust flue gas is a function of all levels air velocities (include A~D primary air, AA~DE secondary air and OFAs ), c) the load of the boiler is the function of A~D speed of powders and the efficiency of the boiler. The real-coded NSGA-II employed by this paper adopted tournament selection method, the simulated binary crossover (SBX) operator, the polynomial mutation operator, and the parameters used include :population size =100, Crossover probability=0.4, Mutation probability=0.05, Maximum number of generation=500. Under the rated load conditions (300MW) and partial loads conditions (200MW), the Pareto-optimal set of multi-objective optimization on the coal-fired boiler obtained by NSGA-II are illustrated fig.5 and fig.6 respectively, and three of Pareto-optimal solutions are shown on tab .1 and tab. 2 respectively, in which the operational parameters are illustrated. Shown as fig.7, the NOx value of Pareto-optimal solution set under the rated load conditions of 300MW varies within the range of 495-630mg/Nm3, the overall heat loss of boiler varies within the range of 8.0-9.05%. Refer to tab.1, NOx emissions from the boiler increases as the Oxygen content supplied to the boiler rises. However, the overall heat loss of boiler decreases as the Oxygen content supplied to the boiler rises, which means a higher efficiency and better economic performance of the coal-fired boiler. That is because with the increase in the amount of oxygen, the reaction of N in the fuel is oxidized into NOx enhanced under the high temperature of fuel gas in the furnace of boiler, and Carbon in the coal is better to burn out, thus the incomplete combustion loss q4 of boiler decreased with the increases of oxygen. Shown as fig.8, the NOx value of Pareto-optimal solution set under the partial loads conditions of 200MW varies within the range of 525-650mg/Nm3, and the overall heat loss of boiler varies within the range of 8.6510.2%. Compared to the 300MW load conditions ,the NOx emission and the overall heat loss of boiler has a different variation law under the partial loads conditions of 200MW , Refer to tab.2, NOx emissions from the boiler decreases as the Oxygen content supplied to the boiler rises. However, the overall heat loss of boiler increases as the oxygen content supplied to the boiler rises, which means a lower efficiency and worse economic performance of the coal-fired boiler. That is because under the low load conditions of boiler combustion, the temperature of furnace is 100-200K
© 2013 ACADEMY PUBLISHER
1305
lower than the rated load of 300MW, with the increase amount of oxygen, the furnace temperature drop lower, the reaction of N in the fuel oxidized into NOx is weaken, under the lower temperature of furnace, Carbon in the coal is more difficult to burn out, thus the incomplete combustion loss q4 of boiler is higher with the increases of oxygen, the exhaust flue gas heat loss of the boiler is increased also as the oxygen contend increased. Compared to the data of literature [3], if the boiler operated under the 300MW load conditions of the Paretooptimal solutions, the overall heat loss can be decreased by 0.3~1.0 percentage , and the NOx emissions can be reduced by 30~40%, in all conditions of the Paretooptimal solutions , NOx emissions from the boiler are all within the national standard of 650mg/Nm3 in China, which demonstrated good results of the NSGA-II applied to the problem of multi-objective optimization of coalfired combustion. V. SUMMARIES The goal of this study is to get a set of Pareto solutions on the problem of multi-objective optimization of coalfired boiler combustions. Modeling and optimizing are two important steps, the first stage of the study was to model the functional relations between outputs (NOx emissions & overall heat loss) and inputs(operational parameters of a utility boiler )by using BP neural network, the predicted NOx emissions and heat loss from the BP neural network model showed good agreement with the measured. In the second stage of study NSGA-II was introduced combined with BP model established above to optimize the operational parameters of the coal-fired boiler, a set of Pareto solutions supplied by NSGA-II algorithm shows good uniformity, the results show that hybrid algorithm by combining BP neural network and NSGA-II can be a good tool to reduce NOx emissions and overall heat loss effectively from coal-fired boiler, that is the method proposed by this paper can simultaneously increase the efficiency of the boiler and reduce the NOx emissions. ACKNOWLEDGMENT This work was supported by National Natural Science Foundation of China (61262048，51167005) and Jiangxi Provincial Office of Education Research Fund (NO: GJJ10293). REFERENCES [1] M. Liu, B. Yi, X.J. Gao, etal. “Nitrogen oxide emissions status of thermal power plants in China and corresponding suggestion,” Environmental Protection, Vol.402, No.8, pp. 7-10.Aug. 2008. [2] GB13223-2003, “Air Pollutant Emissions Standards in Thermal Power Plant of China”, 2003. [3] C. Xu, J.H. Lu, Y. Zhen. “An Experiment and analysis for a Boiler Combustion Optimization on Efficiency and NOx Emissions,” Boiler Technology, Vol.37, No5,pp. 69-74, Oct. 2006. [4] L.G. Liang, Y. Meng, S.L. Wu. “Operation optimization for retrofitted 1025 t/h boiler and experimental study on its
1306
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15] [16]
[17]
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
NOx emission,” Thermal Power Generation, Vol.42,No.1, pp. 63-66, Jan. 2013 Z.Y. GAO, Z. Guo, J.Q. Hu, etal. “Multi-objective combustion optimization and flame reconstruction for W shaped boiler based on support vector regression and numerical simulation”. Proceedings of the CSEE , Vol.31,No5, pp.13-19, May. 2011, H. Zhou, H.B. Zhu, T. H. Zeng , et al. “Artificial neural network modelling on the unburned carbon in fly ash from utility boilers,” Proceedings of the CSEE, Vol.22, No.6, pp. 96-100, Jun. 2002. Ikonen, E., Najim, K., & Kortela, U. “Neuro-fuzzy modeling of power plant flue-gas emissions,”. Engineering Applications of Artificial Intelligence, Vol.13, No.6, pp. 705–717. 2000, P. H. Wang , L. L. Li ,Q. Chen , et al. “Research on applications of artificial intelligence to combustion optimization in a coal-fried boiler,” . Proceedings of the CSEE , Vol.24, No.4, pp. 184-188, Apr. 2004 . C. XU，J. H. Lu ，Y. Zhen, et al, “A Boiler Combustion Global Optimization on Efficiency and Low NOx Emissions Object,” Proceedings of the CSEE , Vol.26, No.4, pp. 46-50,Apr. 2006, F. Gao , Y. J. Zhai, Y. Zhou, et al. “Combustion Optimization for Utility Boilers Based on Sharing LSSVM Model, ”, Journal of Chinese Society of Power Engineering, Vol.32, No.12,pp. 928-933, 2012 C. L. Wang , H. Zhou and G. N. Li , et al. “Support Vector Machine and Genetic Algorithms to Optimize Combustion for Low NOx Emission,” Proceedings of the CSEE, Vol.27,No.11, pp. 40-44,Nov. 2007. T. Y. Mao, Z.Q. Xu and R. Hou, “Efficient constellation design based on improved non-dominated sorting genetic Algorithm-II,” Journal of Computers, Vol.7, No. 6, pp. 1337-1344, 2012 Kaylynmoy Deb, Amrit Partap, Sameer Agarwal, et al. “A fast and elitist multi-objective genetic algorithm：NSGAII, ”. IEEE Transactions on Evolutionary Computation, Vol.6, No.22, pp. 182-197. 2002. J. Y. Zhang, H. W. Zhu, Y. G. Chun , et al. “Multiobjective shape optimization of helico-axial multiphase pump impeller based on NSGA-II and ANN, ”, Energy Conversion and Management, Vol.52, pp. 538-546, 2011. G. L. Zhang and H. Zuo. “Pareto optimal solution analysis of convex multi-objective programming problem, ” Journal of Networks, Vol.8, No.2, pp 437-444, 2013 Adam P. Piotrowski , Jaroslaw J. Napiorkowski, et al. “Differential Evolution algorithm with Separated Groups for multi-dimensional optimization problems,” European Journal of Operational Research ,Vol.216, pp. 33–46, 2012. Horn, J., Nafploitis, N., and Goldberg, D. E. “A niched Pareto genetic algorithm for multi-objective optimization,”
© 2013 ACADEMY PUBLISHER
[18]
[19]
[20] [21]
[22] [23]
Proceedings of the First IEEE Conference on Evolutionary Computation, IEEE Press, New Jersy, pp. 82–87 , 1994 Knowles, J. and Corne, D. “The Pareto archived evolution strategy: a new baseline algorithm for multi-objective optimization,”. Proceedings of the 1999 Congress on Evolutionary Computation, IEEE Press, New Jersy, pp. 98–105, 1999. Kim, M., Hiroyasu, T., Miki, M., and Watanabe, S. “SPEA2+: improving the performance of the strength Pareto evolutionary algorithm 2,”. Computer Science, 3242, 2004, 742–751 F. Q. Gu and H. L. Liu,. “An adaptive multi-objective differential evolution algorithm,” Journal of Computers, Vol.8, No. 2, pp. 294-301, 2013 Chandra, Abhijit, “Role of mutation strategies of differential evolution algorithm in designing hardware efficient multiplier-less low-pass FIR filter, ” Journal of Multimedia, Vol.7, No.5, pp. 353-363, 2012 Srinivas, N. and Deb, K. “Multi-objective optimization using non-dominated sorting in genetic algorithms. Evolutionary Computation”, 1994, 2(3), 221–248 Kalyanmoy Deb, “Multi-Objective Optimization using Evolutionary Algorithms”, First ed., John Wiley & Sons, Ltd., Singapore, 2001, ISBN 9814-12-685-3.
Tingfang Yu received the B. Tech degree in thermal power engineering from Shanghai University of Electric Power, Shanghai, China, 1991 and the M.S. and Ph.D. degrees in thermal power engineering from the Southeast University, Nanjing, China, in 1998 and 2004, respectively. He is currently an Associate Professor of Thermal Power Engineering with the School of Mechanical and electronic Engineering, Nanchang University, Nanchang, China. He has authored or coauthored over 20 research papers in journals and conferences, His current research interests are in the field of coal-fired boiler combustion and optimization algorithm. Hongzhen Zhu received the B. Tech degree in Thermal Power Engineering from, Nanchang University, Nanchang, China, 2010. He is currently a graduate student of Thermal Power Engineering in Nanchang University, Nanchang, China. Chunhua Peng was born in Leping, Jiangxi, China on Oct 15, 1973. He graduated from Southeast University and received doctor degree in 2004. Now, he is an associate professor of Electrical & Electronics Engineering School of East China Jiaotong University, primarily engaged in research of the automation of power system.
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
Multi-objective Optimization of Coal-fired Boiler Combustion Based on NSGA-II Tingfang Yu, Hongzhen Zhu School of Mechanical and Electronic Engineering, Nanchang University, Nanchang, Jiangxi Province, China; Email: [email protected], [email protected]
Chunhua Peng School of Electrical & Electronics Engineering, East China Jiaotong University, Nanchang, Jiangxi Province, China Email: [email protected]
Abstract—NOx emission characteristics and overall heat loss model for a 300MW coal-fired boiler were established by Back Propagation (BP) neural network, by which the the functional relationship between outputs (NOx emissions & overall heat loss of the boiler) and inputs (operational parameters of the boiler) of a coal-fired boiler can be predicted. A number of field test data from a full-scale operating 300MWe boiler were used to train and verify the BP model. The NOx emissions & heat loss predicted by the BP neural network model showed good agreement with the measured. Then, BP model and the non-dominated sorting genetic algorithm II (NSGA-II) were combined to gain the optimal operating parameters which lead to lower NOx emissions and overall heat loss boiler. The optimization results showed that hybrid algorithm by combining BP neural network with NSGA-II can be a good tool to solve the problem of multi-objective optimization of a coal-fired combustion, which can reduce NOx emissions and overall heat loss effectively for the coal-fired boiler. Index Terms—Coal-fired boiler combustion, Multi-objective optimization, BP neural network, NSGA-II
I. INTRODUCTION In China, the requirements for environmental protection are increasingly strict, especially for coal-fired utility boiler. Coal remains the primary energy resource in China, and one of the major concerns associated with coal-fired power plants is the emission of pollutants, especially for NO2 and NO (collectively referred to as NOx). Today, NOx emission is regulated and has become an important consideration in the design and modification of coal-fired utility boiler [1-2]. However, many olddesigned utility boilers in China emit the NOx pollutants above the limit and have posed terrible threat to the surrounding environment, coal-fired power plants face important challenges concerning the methods and technologies to meet these new environmental requirements. In addition to the developments in the plant construction and flue gas cleaners such as Selective Catalytic Reduction (SCR) reactor, NOx control techniques based on combustion modification are of considerable interest[3-5], because they avoid or postpone large capital expenditures while meeting
© 2013 ACADEMY PUBLISHER doi:10.4304/jnw.8.6.1300-1306
environmental compliance requirements compared with the relatively expensive flue gas NOx reduction technologies. The control of the boiler operating conditions through combustion optimization is an important and cost-effective way to affect NOx emissions [3-5]. In recently years, many scholars applied artificial intelligent methods to optimization of coal-fired boiler combustion, H. Zhou [6] established the boiler fly ash carbon content model to predict the relationship between unburned carbon in fly ash and operation parameters of boiler by using BP neural network , Fuzzy neuralnetwork was proposed to model NOx emissions by Ikonen [7]. P. H. Wang [8] established artificial neuralnetworks model of a boiler NOx emissions and efficiency, based on the above model, genetic algorithm (GA) was introduced to get the optimal operation parameters of a coal-fired boiler. C. XU [9] set up the boilers efficiency and NOx emissions model by minimal resource allocating networks (MRAN), simulation studies on global optimization on efficiency and low NOx emissions object were carried out based on MRAN and GA method. A sharing LSSVM model was established by F. Gao [10] to predict the NOx emission and thermal efficiency of a 1000MW boiler, and then adopted an improved PSO algorithm to optimize the operational conditions of the boiler. C. L. Wang [11] using support vector machine (SVM) combined with the ant colony algorithm to optimize the NOx emissions for a coal-fired boiler. However, coal-fired boiler is a very complex system, the level of unburned carbon in fly ash is an important factor affecting the efficiency of pulverized coal fired boiler, and especially those equipped with low NOx burners. Due to the reduced mixing intensity and the formation of fuel rich zones under low NOx combustion conditions, the residence time of the coal particles in the oxygen rich environment decreases, resulting in an increase of the amount of unburned carbon in fly ash. Pollution formation and carbon burnout in pulverized coal combustion are dominated by the fuel properties (reactivity, volatiles, nitrogen content, etc), fuel preparation (coal fineness) and combustion conditions (mixing). The emission of NOx from boiler can be
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
reduced significantly by modification of the combustion process. NOx emission from utility boilers and the boiler efficiency are the different functions of fuel properties, boiler design and operating variables. In the combustion optimization ,the conflict between low NOx emission and high boiler thermal efficiency encounters[4,5], the operation parameters suited to lower NOx emissions of coal-fired boiler always lead to a higher carbon content in fly ash and lower efficiency of the boiler, the above studies on optimization of coal-fired boiler combustion mainly focus on the single-objective optimization ,for only on the emissions of NOx from the boilers or only boiler efficiency alone, and can’t reach both lower NOx emissions and higher efficiency of boiler ,It is imperative to find a good tool for multi-objective optimization of coal-fired boiler combustion. The traditional approach is converting the multiobjective problem(MOP) into single objective problem(SOP) by means of assign different weights to all objectives , then get the optimal point by means GA or other algorithms as in the literature[8,9], its disadvantage include: a) it can’t get a set of Pareto optimal solutions but a single solution, b)different objectives is not comparable, thus a reasonable weights does not exist, the assigning of weights introduced human factors inevitably ,which may be debatable. The real result of multi-objective problem solving should be a set of Pareto optimal solutions instead of just to get only an optimal solution [12-14]. There is no single solution that is optimal (global minimum or maximum) with respect to all the optimization objectives under multi-constraints for the multi-objective optimization contrasted to the singleobjective one, and only acceptable non-dominated solutions exist, which are the so-called Pareto optimal solutions[15,16]. The approaches for solving conventional multi-objective problems include the objective weighted method, the hierarchical optimization method, the goal programming method, etc. These algorithms deal with the problems converting the multiobjective problem to the single-objective problem with the emphasis of one particular Pareto-optimal solution at a time. Actually, they are not real multi-objective optimization algorithms and cannot be widely used in solving the complex problems. Over the past decades, to solve the complicated multi-objective optimization problems and achieve the Pareto optimal solution set, a number of multi-objective evolutionary algorithms have been suggested, such as the niched Pareto genetic algorithm (NPGA)[17], the Pareto archived evolution strategy (PAES)[18], the strength Pareto evolutionary algorithm(SPEA)[19], Differential Evolution algorithm[20,21] and the non-dominated sorting genetic algorithm (NSGA)-II[12-14]. Among them, the recently well-known NSGA-II has drawn more attention due to its feasibility for any number of objectives. In this paper, in order to get the minimum NOx emissions and maximum boiler efficiency, taking the minimum the overall heat loss of the boiler and the minimum NOx emissions as the optimization objectives.
© 2013 ACADEMY PUBLISHER
1301
a BP neural-network were proposed to model NOx emissions and overall heat loss of the coal-fired boiler ,then the NSGA-II is introduced as multi-objective optimization algorithms combined with BP neural network model to obtain the Pareto optimal solution set of the multi-objective optimization of coal-fired boiler combustion problem. The rest of the paper is organized as follows: Section II describes mathematical model of NOx emissions and boiler heat loss established by BP neural network based on the test data of a 300MWe coal-fired boiler, the simulation results of BP model were given, and Analyzed the accuracy of the model. Section III introduced NSGAII algorithm and steps involved in implementation to the multi-objective optimization of a coal-fired boiler combustion problem. Section IV performed application of NSGA-II to multi-objective optimization of coal-fired boiler combustion, numerical results were given under the rated load of 300MW and a partial load of 200MW, Pareto-optimal set of multi-objective optimization on the coal-fired boiler obtained by NSGA-II were illustrated, the NOx emission and the overall heat loss of boiler shown a different variation law under the partial loads conditions and rated load, then the reason for the above laws were illustrated , finally in Section V summaries are given. II. MATHEMATICAL MODEL A. Description of the Boiler The boiler is a 300 MW, tangentially fired dry bottom boiler with double furnaces, equipped with indirect-fired coal-storage pulverizing system and exhaust pneumatic convey system, four DTM350/600 drum ball mill, four exhaust fans,32 powders, using horizontal and louver type concentrated diluted pulverized coal burner. The maximum continuous rating (MCR) of the boiler is 1025 t/h of superheated steam at 538℃. The boiler had four levels of primary air nozzles (A, B, C, D) and five levels of secondary air nozzles (AA, AB, BC, CD, DE), two levels of over-fire air nozzles(OFA1,OFA2), a level separated over-fire air nozzles(SOFA). The Cross-section of the boiler furnace and the arrangement of the boiler burner nozzles are shown as figure 1 and figure 2
Figure 1. Cross-section of the boiler furnace
1302
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
neuron in the output layer, according to a large number of simulation , we finally select 21 neurons in the hidden layer for the BP model show as in fig.3.
Predict NOx values by BP model/10-4 mg/m3
C. Simulation Results Training the BP neural network is an important step for developing a useful network. On the basis of 20 test data of the 300MWe boiler in the literature [3], the experimental data of Cases 1–18 are used as the learning samples to train the BP neural network. 2 others are employed as input to verify the accuracy of the model. The moment attachment and the self-adaptive learning step size method are adopted in the training process. The momentum factor is taken as 0.5. The training will stop when the mean square error of the system is less than 106, a total of 20000 epochs are needed to achieve the correct weight and threshold values.
Figure 2. Arrangement of the boiler burner bozzles
Velocities of primary air (4 variables) Speed of four levels powders (4 variables)
NOx
Opening of OFA1/OFA2/SOFA (3 variables) Velocities of secondary air (5 variables) Oxygen contend(1 variable)
Heat loss
9.5
Training samples Verify samples
9
8.5
8
7.5
7
6.5
6
5.5
5
BP neural network mathematical model of the coal-fired boiler combustion characteristics
B. BP Model of NOx Emissions and Boiler Heat Loss As BP neural network has good nonlinear mapping ability, the paper introduced the BP neural network to establish mathematical model of the coal-fired boiler combustion characteristics, which is used to evaluate the NOx emissions (at 6 vol%O2 dry) from the boiler and the overall heat loss of boiler. Moreover, the model can be employed as the objective functions in the later multiobjective optimization of coal-fired boiler combustion. On the basis of test data of the 300MWe boiler[3], BP neural network mathematical model of the coal-fired boiler combustion characteristics under the steady-state operation conditions was established, 17 parameters closely related to the NOx emissions and the overall heat loss from the boiler were selected as the input variables: four levels primary air average velocities(A-D), four levels average speed of the powders(A-D), five levels secondary air average velocities(AA-DE), three opening percent of the over-fire air nozzles (OFA1/OFA2/SOFA) and the oxygen content in the exhaust gas from the boiler. The output of the BP model are NOx emissions and the overall heat loss from the boiler, thus the BP neural network has 17 neurons with input layer, two output
© 2013 ACADEMY PUBLISHER
5
Figure 4.
5.5
6
6.5
7
7.5
8
8.5
9
9.5
the measured NOx values /10-4
Simulation results of the NOx emissions from the boiler by BP model
11.5
11
Predict values by BP model /%
Figure 3.
Training samples Verify samples
10.5
10
9.5
9
8.5
8
7.5 7.5
8
8.5
9
9.5
10
10.5
11
11.5
The measured overall heat loss values/% Figure 5. . Simulation results of the boiler heat loss by BP model
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
After the BP model training was completed, the simulation results of the NOx emissions from the boiler by BP model are shown in figure 4, and simulation results of the boiler heat loss by BP model are illustrated in fig.5. The figure shows that BP neural network model predict the NOx emissions from the boiler are not more than 1.8% deviation compared with the measured values, and the overall heat loss are not more than 2.3%, which can satisfied the accuracy requirement of the engineering. III. PROPOSED SCHEME A. Description of the NSGA-II The NSGA proposed by Srinivas and Deb [22] is an efficient sorting algorithm and developed from the simple genetic algorithm only in the way of the selection operator. The NSGA can simplify multi-objective optimization problems into one fitness function problem and also be used for any objective of optimization problems. The main ideas of the NSGA are as follows: First, the individuals of the population must be classified into different non-dominated sets according to the nondomination relationship. Second, to maintain the diversity in the population, sharing methods are adopted to control the individual distribution of every non-dominated set. Third, the selection operator is performed on the basis of an individual’s domination set and niche count. Finally, combined with other operators of the simple genetic algorithm, the Pareto-optimal solutions can be obtained. Deb et al. [13] carried out a substantial improvement of the NSGA and proposed a fast and elitist NSGA (i.e., NSGA-II) with the following three aspects of improvement: (i) A fast non-dominated sorting approach taken to reduce the computational complexity from O(MN3) to O(MN2), where M is the number of the objectives, and N is the population size, (ii) The elitism strategy introduced to improve the robustness and convergence, (iii) The sharing methods replaced by the crowdingdistance methods to avoid specifying the sharing parameter δ share. Combined with the tournament selection method, the simulated binary crossover (SBX) operator, the polynomial mutation operator, and the efficient constraint-handling approach, the NSGA-II displays good convergence and robustness [13]. 1) The SBX crossover operator SBX crossover operator was adopted in NSGA-II, cross-point position was received randomly, and the genetic codes of two parent individuals on both sides of the crossing point were exchanged. In general, SBX crossover puts the stress on generating offspring near the parents. This crossover guarantees that the extent of the children or offspring is proportional to the extent of the parents, and also favors that near parent individuals are monotonically more likely to be chosen as children than individuals distant from the parents in the solution space. The procedure for finding the offspring solutions from parent solutions is given in [13]. © 2013 ACADEMY PUBLISHER
1303
2) Polynomial mutation The probability of creating a solution yit+1 near to the parent is higher than the probability of creating one distant from it. The shape of the probability distribution is directly controlled by an external parameterηm and the distribution remains unchanged throughout the iterations. Like in the SBX operator, the probability distribution can also be a polynomial function, instead of a normal distribution [23]: yit+1=xit+1+(xi(U)−xi(L) )δi
(1)
where the parameterδi is calculated from the polynomial probability distribution, xit+1 and xi are one of the offspring solutions and parent solutions respectively, superscript U and L represents the upper and lower limit for the corresponding variables. P(δi)=0.5(ηm +1)(1−|δi |)ηm
δi
=(2ri)1/ηm −1 −1, if ri < 0.5 =1−[2(1−ri)]1/ηm +1, if ri ≥0.5
(2)
For handling the bounded decision variables, the mutation operator is modified for two regions: [xi(L), xi] and [xi, xi(U)] [23]. B. Algorithm Implementation Fig. 6 explains the implementation of NSGA-II algorithm for the multi-objective coal-fired boiler combustion problem. It is described in the following steps: Step 1: Identify the control variables limits of the boiler operational parameters like velocities of primary air, speed of powders, oxygen contend and velocities of secondary air for the multi-objective optimization of a coal-fired boiler combustion problem. Step 2: Select the parameters like number of population, maximum number of generation, crossover and mutation probabilities. Step 3: Generate initial population. Step 4: Evaluate objective functions (i.e. NOx and Heat loss of the boiler) by the BP model in section II for initial population. Step 5: Set the generation count. Step 6: Perform the SBX crossover operator and polynomial mutation for the set of individuals. Step 7: Perform non-dominated sorting [13] (i.e., sorting the population according to each objective function value in ascending order of magnitude). Step 8: Generate population for next generation from combined parent and offspring population. Step 9: Perform selection based on tournament selection [13] thereby a higher fitness is assigned to individuals located on a sparsely populated part of the front. Step 10: Increment the generation count and repeat the steps from 6 to 9 until the count reaches the maximum number of generation.
1304
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
6.4
NOx emissions form the boiler 10-4 mg/N3)
Identify the control variables limits of velocities of primary air, speed of powders，oxygen contend and velocities of secondary air
Select the parameters like population size, maximum number of generation, crossover and mutation probabilities
Generate initial population
Evaluate objective functions (NOx and Heat loss of boiler) for initial population
Set the generation count, i = 0
6.3 6.2 6.1 6 5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 5.1 5
Generate offspring population using SBX crossover operator and
4.9 8
polynomial mutation
8.1
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
9
9.1
The overall heat loss of boiler (%)
Perform non-dominated sorting for combined parent population and offspring population
Figure 7. Pareto-optimal solution set of multi-objective optimization of a coal-fired boiler combustion (300MW) i=i+1
Generate population for next generation from combined parent and offspring population
6.8
Perform parent selection based on tournament selection
Stop
i < i max
Figure 6. Flowchart of NSGA-II algorithm for the multi-objective coal-fired boiler combustion problem.
IV. NUMERICAL RESULTS The optimization objectives of the multi-objective optimization of coal-fired boiler combustion include the minimum of NOx emissions from boiler and the maximum of the boiler efficiency (100%-overall heat loss%), thus taking the minimum the overall heat loss of the boiler and the minimum NOx emissions as the optimization objectives in this paper.
NOx emissions form the boiler 10-4 mg/N3)
6.7 6.6 6.5 6.4 6.3 6.2 6.1 6 5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 5.1 8.6 8.7 8.8 8.9
9
9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 10 10.1 10.2
The overall heat loss of boiler (%)
Figure 8. Pareto-optimal solution set of multi-objective optimization of a coal-fired boiler combustion (200MW)
TABLE I. . OPERATION CONDITIONS OF THREE PARETO SOLUTIONS(300MW)
a b c
Primary air velocity (m/s) A B C D 28.2 28.5 28.6 29 28.6 28.5 29.1 29.3 29.2 28.9 28.9 29.5
Speed of powders(rpm) A B C D 437 402 339 335 442 395 353 334 446 392 351 322
TABLE II.
a b c
Primary air velocity (m/s) A B C D 25.5 26.1 25.8 26.2 26.7 26.5 26.6 27.1 28.1 28.9 29.5 28.8
© 2013 ACADEMY PUBLISHER
Secondary air velocity (m/s) Opening of OFA(%) Oxygen Heat loss AA AB BC CD DE OF1 OF2 SOFA % % 30.1 32.6 31.2 31.8 31.6 73 75 55 4.03 9.02 31.3 31.7 32.3 32.2 32.5 86 85 61 4.65 8.47 32.3 32.8 33.6 33.8 34.7 92 95 82 5.12 8.02
NOx 10-4 5.01 5.55 6.35
. OPERATION CONDITIONS OF THREE PARETO SOLUTIONS (200MW)
Speed of powders(rpm) A B C D 437 386 329 315 442 365 323 321 449 360 324 312
Secondary air velocity (m/s) Opening of OFA(%) Oxygen Heat loss AA AB BC CD DE OF1 OF2 SOFA % % 27.2 26.6 26.2 28.0 20.6 43 45 0 5.06 8.73 27.5 27.9 28.1 29.7 22.7 48 51 0 5.84 9.55 28.6 29.8 31.3 30.9 21.5 52 56 0 6.65 10.15
NOx 10-4 6.48 5.65 5.39
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
In the process of optimization with the NSGA-II algorithm, the premises include maintain the stability of the boiler load, the same coal injected into the boiler. Moreover, In order to ensure optimal accuracy and reasonableness of the optimization results ,the are also some constraints must be taken into account, those constraints include: a) Upper and lower limits of the optimization variables, the range of optimization variables are: primary air velocity of 25~30m/s, the secondary air in the AA~DE levels were 25~40m/s, OFA dampers opening as 0~100%, oxygen to optimize rang of 3% to 6.5%, the speed of powders ranges from 300~500rpm. b) Oxygen in the exhaust flue gas is a function of all levels air velocities (include A~D primary air, AA~DE secondary air and OFAs ), c) the load of the boiler is the function of A~D speed of powders and the efficiency of the boiler. The real-coded NSGA-II employed by this paper adopted tournament selection method, the simulated binary crossover (SBX) operator, the polynomial mutation operator, and the parameters used include :population size =100, Crossover probability=0.4, Mutation probability=0.05, Maximum number of generation=500. Under the rated load conditions (300MW) and partial loads conditions (200MW), the Pareto-optimal set of multi-objective optimization on the coal-fired boiler obtained by NSGA-II are illustrated fig.5 and fig.6 respectively, and three of Pareto-optimal solutions are shown on tab .1 and tab. 2 respectively, in which the operational parameters are illustrated. Shown as fig.7, the NOx value of Pareto-optimal solution set under the rated load conditions of 300MW varies within the range of 495-630mg/Nm3, the overall heat loss of boiler varies within the range of 8.0-9.05%. Refer to tab.1, NOx emissions from the boiler increases as the Oxygen content supplied to the boiler rises. However, the overall heat loss of boiler decreases as the Oxygen content supplied to the boiler rises, which means a higher efficiency and better economic performance of the coal-fired boiler. That is because with the increase in the amount of oxygen, the reaction of N in the fuel is oxidized into NOx enhanced under the high temperature of fuel gas in the furnace of boiler, and Carbon in the coal is better to burn out, thus the incomplete combustion loss q4 of boiler decreased with the increases of oxygen. Shown as fig.8, the NOx value of Pareto-optimal solution set under the partial loads conditions of 200MW varies within the range of 525-650mg/Nm3, and the overall heat loss of boiler varies within the range of 8.6510.2%. Compared to the 300MW load conditions ,the NOx emission and the overall heat loss of boiler has a different variation law under the partial loads conditions of 200MW , Refer to tab.2, NOx emissions from the boiler decreases as the Oxygen content supplied to the boiler rises. However, the overall heat loss of boiler increases as the oxygen content supplied to the boiler rises, which means a lower efficiency and worse economic performance of the coal-fired boiler. That is because under the low load conditions of boiler combustion, the temperature of furnace is 100-200K
© 2013 ACADEMY PUBLISHER
1305
lower than the rated load of 300MW, with the increase amount of oxygen, the furnace temperature drop lower, the reaction of N in the fuel oxidized into NOx is weaken, under the lower temperature of furnace, Carbon in the coal is more difficult to burn out, thus the incomplete combustion loss q4 of boiler is higher with the increases of oxygen, the exhaust flue gas heat loss of the boiler is increased also as the oxygen contend increased. Compared to the data of literature [3], if the boiler operated under the 300MW load conditions of the Paretooptimal solutions, the overall heat loss can be decreased by 0.3~1.0 percentage , and the NOx emissions can be reduced by 30~40%, in all conditions of the Paretooptimal solutions , NOx emissions from the boiler are all within the national standard of 650mg/Nm3 in China, which demonstrated good results of the NSGA-II applied to the problem of multi-objective optimization of coalfired combustion. V. SUMMARIES The goal of this study is to get a set of Pareto solutions on the problem of multi-objective optimization of coalfired boiler combustions. Modeling and optimizing are two important steps, the first stage of the study was to model the functional relations between outputs (NOx emissions & overall heat loss) and inputs(operational parameters of a utility boiler )by using BP neural network, the predicted NOx emissions and heat loss from the BP neural network model showed good agreement with the measured. In the second stage of study NSGA-II was introduced combined with BP model established above to optimize the operational parameters of the coal-fired boiler, a set of Pareto solutions supplied by NSGA-II algorithm shows good uniformity, the results show that hybrid algorithm by combining BP neural network and NSGA-II can be a good tool to reduce NOx emissions and overall heat loss effectively from coal-fired boiler, that is the method proposed by this paper can simultaneously increase the efficiency of the boiler and reduce the NOx emissions. ACKNOWLEDGMENT This work was supported by National Natural Science Foundation of China (61262048，51167005) and Jiangxi Provincial Office of Education Research Fund (NO: GJJ10293). REFERENCES [1] M. Liu, B. Yi, X.J. Gao, etal. “Nitrogen oxide emissions status of thermal power plants in China and corresponding suggestion,” Environmental Protection, Vol.402, No.8, pp. 7-10.Aug. 2008. [2] GB13223-2003, “Air Pollutant Emissions Standards in Thermal Power Plant of China”, 2003. [3] C. Xu, J.H. Lu, Y. Zhen. “An Experiment and analysis for a Boiler Combustion Optimization on Efficiency and NOx Emissions,” Boiler Technology, Vol.37, No5,pp. 69-74, Oct. 2006. [4] L.G. Liang, Y. Meng, S.L. Wu. “Operation optimization for retrofitted 1025 t/h boiler and experimental study on its
1306
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15] [16]
[17]
JOURNAL OF NETWORKS, VOL. 8, NO. 6, JUNE 2013
NOx emission,” Thermal Power Generation, Vol.42,No.1, pp. 63-66, Jan. 2013 Z.Y. GAO, Z. Guo, J.Q. Hu, etal. “Multi-objective combustion optimization and flame reconstruction for W shaped boiler based on support vector regression and numerical simulation”. Proceedings of the CSEE , Vol.31,No5, pp.13-19, May. 2011, H. Zhou, H.B. Zhu, T. H. Zeng , et al. “Artificial neural network modelling on the unburned carbon in fly ash from utility boilers,” Proceedings of the CSEE, Vol.22, No.6, pp. 96-100, Jun. 2002. Ikonen, E., Najim, K., & Kortela, U. “Neuro-fuzzy modeling of power plant flue-gas emissions,”. Engineering Applications of Artificial Intelligence, Vol.13, No.6, pp. 705–717. 2000, P. H. Wang , L. L. Li ,Q. Chen , et al. “Research on applications of artificial intelligence to combustion optimization in a coal-fried boiler,” . Proceedings of the CSEE , Vol.24, No.4, pp. 184-188, Apr. 2004 . C. XU，J. H. Lu ，Y. Zhen, et al, “A Boiler Combustion Global Optimization on Efficiency and Low NOx Emissions Object,” Proceedings of the CSEE , Vol.26, No.4, pp. 46-50,Apr. 2006, F. Gao , Y. J. Zhai, Y. Zhou, et al. “Combustion Optimization for Utility Boilers Based on Sharing LSSVM Model, ”, Journal of Chinese Society of Power Engineering, Vol.32, No.12,pp. 928-933, 2012 C. L. Wang , H. Zhou and G. N. Li , et al. “Support Vector Machine and Genetic Algorithms to Optimize Combustion for Low NOx Emission,” Proceedings of the CSEE, Vol.27,No.11, pp. 40-44,Nov. 2007. T. Y. Mao, Z.Q. Xu and R. Hou, “Efficient constellation design based on improved non-dominated sorting genetic Algorithm-II,” Journal of Computers, Vol.7, No. 6, pp. 1337-1344, 2012 Kaylynmoy Deb, Amrit Partap, Sameer Agarwal, et al. “A fast and elitist multi-objective genetic algorithm：NSGAII, ”. IEEE Transactions on Evolutionary Computation, Vol.6, No.22, pp. 182-197. 2002. J. Y. Zhang, H. W. Zhu, Y. G. Chun , et al. “Multiobjective shape optimization of helico-axial multiphase pump impeller based on NSGA-II and ANN, ”, Energy Conversion and Management, Vol.52, pp. 538-546, 2011. G. L. Zhang and H. Zuo. “Pareto optimal solution analysis of convex multi-objective programming problem, ” Journal of Networks, Vol.8, No.2, pp 437-444, 2013 Adam P. Piotrowski , Jaroslaw J. Napiorkowski, et al. “Differential Evolution algorithm with Separated Groups for multi-dimensional optimization problems,” European Journal of Operational Research ,Vol.216, pp. 33–46, 2012. Horn, J., Nafploitis, N., and Goldberg, D. E. “A niched Pareto genetic algorithm for multi-objective optimization,”
© 2013 ACADEMY PUBLISHER
[18]
[19]
[20] [21]
[22] [23]
Proceedings of the First IEEE Conference on Evolutionary Computation, IEEE Press, New Jersy, pp. 82–87 , 1994 Knowles, J. and Corne, D. “The Pareto archived evolution strategy: a new baseline algorithm for multi-objective optimization,”. Proceedings of the 1999 Congress on Evolutionary Computation, IEEE Press, New Jersy, pp. 98–105, 1999. Kim, M., Hiroyasu, T., Miki, M., and Watanabe, S. “SPEA2+: improving the performance of the strength Pareto evolutionary algorithm 2,”. Computer Science, 3242, 2004, 742–751 F. Q. Gu and H. L. Liu,. “An adaptive multi-objective differential evolution algorithm,” Journal of Computers, Vol.8, No. 2, pp. 294-301, 2013 Chandra, Abhijit, “Role of mutation strategies of differential evolution algorithm in designing hardware efficient multiplier-less low-pass FIR filter, ” Journal of Multimedia, Vol.7, No.5, pp. 353-363, 2012 Srinivas, N. and Deb, K. “Multi-objective optimization using non-dominated sorting in genetic algorithms. Evolutionary Computation”, 1994, 2(3), 221–248 Kalyanmoy Deb, “Multi-Objective Optimization using Evolutionary Algorithms”, First ed., John Wiley & Sons, Ltd., Singapore, 2001, ISBN 9814-12-685-3.
Tingfang Yu received the B. Tech degree in thermal power engineering from Shanghai University of Electric Power, Shanghai, China, 1991 and the M.S. and Ph.D. degrees in thermal power engineering from the Southeast University, Nanjing, China, in 1998 and 2004, respectively. He is currently an Associate Professor of Thermal Power Engineering with the School of Mechanical and electronic Engineering, Nanchang University, Nanchang, China. He has authored or coauthored over 20 research papers in journals and conferences, His current research interests are in the field of coal-fired boiler combustion and optimization algorithm. Hongzhen Zhu received the B. Tech degree in Thermal Power Engineering from, Nanchang University, Nanchang, China, 2010. He is currently a graduate student of Thermal Power Engineering in Nanchang University, Nanchang, China. Chunhua Peng was born in Leping, Jiangxi, China on Oct 15, 1973. He graduated from Southeast University and received doctor degree in 2004. Now, he is an associate professor of Electrical & Electronics Engineering School of East China Jiaotong University, primarily engaged in research of the automation of power system.
