RESEARCH ON DESIGN OPTIMIZATION OF ENERGY SOURCE SIZES AND CONTROL PARAMETERS OF HYBRID ELECTRIC VEHICLE POWERTRAIN SYSTEMS

MINISTRY OF EDUCATION AND TRAINING NHA TRANG UNIVERSITY VU THANG LONG RESEARCH ON DESIGN OPTIMIZATION OF ENERGY SOURCE SIZES AND CONTROL PARAMETERS OF HYBRID ELECTRIC VEHICLE POWERTRAIN SYSTEMS Mayor : Transportation Engineering Major Code : 62520116 DOTOCRAL DISSERTATION SUMMARY Khanh Hoa – 2015 Research was accomplished at Nha Trang University Supervisor: Prof Nguyen Van Nhan Referee : Referee : Referee : The thesis was defended at Nha Trang university committee of doctorate thesis examiners at … The disseration can be found at … INTRODUCTION The rationale Hybrid electric vehicles (HEVs) are one of transient solutions to solve fuel save requirements and environmental polulltion caused by emissions of internal combustion engine (ICE) The optimization method of key component sizes and control strategy parameters of HEV powertrain system is one of prerequisite conditions to design and fabricate HEVs Study on theory and hybridisation applications in order to base for the design and fabrication of HEVs and evaluate the advantages of HEVs using in Vietnamses road conditions, the research topic “Research on design optimization of energy source sizes and control strategy parameters of hybrid electric vehicle powertrain systems” has been carried out Research Objectives Building a simultaneous multi-objective optimization model of energy source sizes and control strategy parameters of HEV powertrain systems in order to reduce fuel consumption (FC) and improve toxic emissions of ICE Research Objects Hybrid vehicles have a powertrain system composed from two kinds of power sources including an ICE and electric motors (EMs) Research Scope (1) All energy sources in a general configuration of powertrain systems of parallel HEVs and parallel-series HEVs, including ICE, EM, electric generator (EG) and battery (AQ) (2) Design optimization of powertrain energy source sizes of parallel and seriesparallel HEVs (3) Optimization of energy source control parameters of parallel and seriesparallel HEV powertrain Reseach Contents The research contents include: HEV overview and optimization of HEV powertrain systems; Application of a Bees Algorithm as an optimizer for powertrain component sizing and control strategy parameter optimization for hybrid electric vehicles; Experimental simulation of key component size and control strategy parameter optimization for powertrain system; Conclusions and recomendations Limitations The dissertation mainly focuses on the powertrain component sizing and control strategy parameter optimization for hybrid electric vehicles in order to reduce fuel consumption and emissions of ICE without sacrificing on-road performances The design optimization method of energy source sizes and control strategy parameters of HEV powertrain systems by using bees algorithms has not been experimented on physical HEVs in laboratory or in real operating conditions Key Findings (1) Development of multi-objective optimization model of energy source sizes and control strategy parameters of hybrid electric vehicle powertrain systems; (2) Developing equations fs(Sj) (2.30), nb(Sj,t) (2.32) and ph(Sj,t) (2.34) of pheromone-based Bees Algorithm; (3) Building penalty functions to optimize energy source sizes and control strategy parameters of hybrid electric vehicles (4) Building a software module to optimize energy source sizes and control strategy parameters of hybrid electric vehicles by means of ADVISOR software in which Vietnamese dynamic performances and driving cycle are taken into account Chapter OVERVIEW OF HYBRID ELECTRIC VEHICLES AND HYBRID ELECTRIC VEHICLE POWERTRAIN SYSTEM OPTIMIZATION Design optimization of energy source sizes of hybrid electric vehicle powertrain system: The HEV powertrain system has at least two kinds of different power sources including an ICE and an or more EMs They play the main role of providing kinematic energy for driving wheels, however, the powertrain needs to have a battery system and an EG for storing and producing electric energy The size of ICE and EM means maximum power The size of battery system is its total capacity Design optimization of energy source sizes of hybrid electric vehicle powertrain systems is the determination of maximum suitable power of ICE, EM, EG and battery capacity to satisfy optimization objectives Optimization of control strategy parameters of hybrid electric vehicle powertrain systems: That is a working mode determination and supervision of energy sources to get optimization objectives of HEV inherent advantages Most of researches on HEV optimization focus on : - Design and fabrication of HEV models and some of key components of HEV powertrain systems - Research on a control strategy parameter optimization: Particle swarm optimization (PSO) method; Genetic algorithm (GA) method and Simulated annealing (SA) method - Research on design optimization of power source sizes: Power source size optimization for series HEVs using a combination between GA and Sequential Quadratic Programming (SQP); Power source size optimization for parallel HEVs using three algorithms including DIRECT algorithm, SA and GA to improve FC - Research on a simultaneous optimization of key power source sizes and control strategy parameters: Optimization for series and series-parallel HEVs using GA Above analyses and literature reviews about HEVs show that most of domestic researches focus on design and fabrication of HEV models and some key components of HEV powertrain systems with the purpose of deeply understanding about HEVs or serving for training education Besides, external researches published in journals focus on separate optimizations of power source sizes or only control strategy parameters Some latest researches have carried out a simultaneous optimization of power source sizes and control strategy parameters using GA However, since the inherent characteristic of GA is based on a natural selection through many generations, so the convergence speed is very slow, and these researches have not been applied to optimize for all parameters of HEV powertrain systems, some parameters are still kept in default values From existing problems in researching on HEVs, in this research, new methods will be used to optimize energy source sizes and control strategy parameters of HEV powertrain systems Bees algorithm is chosen and developed to solve parallel and series-parallel HEV optimization ADVISOR and Matlab software will be applied to simulate optimization experiments for Honda Insight 2000 and Toyota Prius 1998, in which Vietnamese constraint conditions are also taken into account Chapter APPLICATION OF A BEES ALGORITHM AS AN OPTIMIZER FOR POWERTRAIN COMPONENT SIZING AND CONTROL STRATEGY PARAMETER OPTIMIZATION FOR HEV POWERTRAIN SYSTEMS Chapter presents a general optimization model of energy source sizes and control strategy parameters of HEV powertrain systems composed from an ICE and EMs Basic characteristics of BA, some developed innovations about BAs and its applications for only control strategy parameter optimization and simultaneous optimization of control strategy parameters and energy source sizes are fundamental theories of experimental simulations presented in chapter A GENERAL OPTIMIZATION MODEL OF ENERGY SOURCE SIZES AND CONTROL STRATEGY PARAMETERS OF HEV POWERTRAIN SYSTEMS Min  G( X )   Min( FC, HC, CO, NO) hl ( X )   k q ( X )  OPTIMIZATION UNIT X Y = F(X) DRIVING CYCLE EQUIPTMENTS / HEV MODEL CONTROL STRATEGY Fig 2-1: A general optimization model for HEVs using MDO The author has been used Multidisciplinary Design Optimization (MDO) to solve a problem of multi-parameter optimization of HEV powertrain systems composed from components having different structure characteristics and operating principles as shown in Fig 2-1 G ( X )  w1.FC  w2 HC  w3 CO  w4 NO w  w  w  w    hl ( X )  kq ( X )    X CZ 1 xCZ 1 _  xCZ 1 _ max        X CZ  n xCZ  n _  xCZ  n _ max    X   X CS 1 xCS 1 _  xCS 1 _ max        X CS  m xCS  m _  xCS  m _ max    (2.1) (2.2)  FC ( X )   HC ( X )   Y  CO ( X )     NO ( X )  (2.3) Where : G(X) is objective function hl(X) ≤ và kq(X) ≥ are constraints about HEV dynamic performance characteristics XCZ-i (i = ÷ n) is a variable set of key component sizes of powertrain system including maximum power of ICE, EM, EG and battery capacity XCS-j (j = ÷ m) is a variable set of control strategy parameters Y is a output set including FC, contents of HC, CO and NO In this dissertation, an oriented trial and error method is used for the optimization process of energy source sizes and control strategy parameters of HEV powertrain systems This process includes many iterations, at each iteration, OPTIMIZATION UNIT will assign a specific value for each variable of X in a range bounded by lower and upper limits as shown in Eq (2.2) to create a “optimal candidate” Then, the candidate will be sent to the modeled EQUIPMENT block Control strategy will determine an operating point of all power sources according to each driving cycle period The output Y of EQUIPTMENT block will be FC, contents of HC, CO and NO in ICE emissions, and HEV dynamic performances All components of Y will be returned to OPTIMIZATION UNIT to calculate G(X) and check dynamic performence constraint requirements Basing on results of G(X) and checking constraints, OPTIMIZATION UNIT will change component values of X through an optimization algorithm to create a “new candidate” The process will be repeated until optimization conditions are satisfied Next part, innovations and applications of BAs as an OPTIMIZATION UNIT to optimize key component sizes and control strategy parameters of HEV powertrain systems BASIC BEES ALGORITHM (BBA) The optimization process for HEVs using BBA is described as follows:  Step 1: Initialize the population of n scout bees, each scout bee is a set of specific value of all variables of key component sizes and control strategy parameters  Step 2: Evaluate the FC, HC, CO, NO and penalty functions Ci(x) for each scout bee by combining between BBA and ADVISOR software  Step 3: Calculate the fitness value of all scout bees according to Eq (2.22) and choose m scout bees with ranked fitness from the highest to lowest value  Step 4: Conduct n1 searches in each neighbourhood of the best e sites and choose the bee with highest fitness for each site  Step 5: Conduct n2 searches in each neighbourhood of the (m-e) sites and choose the bee with highest fitness at each site  Step 6: Assign remaining (n-m) bees to search randomly around the search space for new potential solutions These searches are carried out to avoid a local optimization  Step 7: Update new population from best bees of m and (n-m) sites  Step 8: Stop the program if the convergence criteria is satisfied, otherwise go to step The convergence criterion is that the fitness of the best bee of the new population is not improved after the specific number of iterations or after N iterations In order to apply BA to the simultaneous optimization of HEVs, the fitness in step is the inverse of objective function G(X) in Eq (2.1) However, the optimization task is required to maintain Vietnamese dynamic performances according to TCVN 4054 : 2005 22 TCN 307 – 03 Unfortunately, BBA can not work directly with constrained optimization problem To solve this problem, it is necessary to add penalty functions into objective function G(X) as shown from Eq (2.22) to (2.24) The penalty functions are used to penalize the infeasible solutions by reducing their fitness values Ci(X) = 0, if its constrain is satisfied Fn( X )  G ( X )   ki Ci ( X )  G (X ) ' (2.22) i 1 G ' ( X )  G( X )   ki Ci ( X ) (2.23) C1 ( X )  max(0; vmax  Velocity ( X )) / vmax  C2 ( X )  max(0;  Slope( X )) /  C ( X )  max(0; Time( X )  t ) / t 200 m 200 m  (2.24) i 1 Where : X = (x1, x2, …, xn) are parameters of key component sizes and control strategy Ci(X), αi and Fi(X) are penalty function, desired value and evaluated value related to ith constrain ki is penalty factor chosen by a trial and error method G ' ( X ) is objective function in which constraint requirements are taken into account vmax , θmin t200m are maximum velocity, minimum grability and necessary time to run from to 200m PHEROMONE-BASED BEES ALGORITHM Similar to BBA, the fitness function Fn(X) is also an invert of G’(X) The penalty function at Eq (2.28) is added into Eq (2.27) to consider dynamic constraint requirements of HEVs The value of fitness at site Sj is calculated as follows : fitness( S j )( x )  G ( S j )( x )   ki Ci ( S j )( x )  G ( S j )( x ) ' (2.27) i 1 C1 ( S j )( x )  max(0; vmax  Velocity ( S j )( x ) ) / vmax  C2 ( S j )( x )  max(0;  Slope( S j )( x ) ) /   C3 ( S j )( x )  max(0; Time( S j )( x )  t200 m ) / t200 m (2.28) The optimization process for HEVs using PBA is described as follows:  Step 1: Initialize the population of n scout bees, each scout bee is a set of specific values of all variables of key component sizes and control strategy parameters  Step 2: Evaluate the FC, HC, CO, NO and penalty functions Ci(X) for each scout bee by combining between PBA and ADVISOR software  Step 3: Calculate the fitness value of all scout bees according to Eq (2.27), (2.28) and (2.30)  Step 4: Choose e bees with highest fitness  Step 5: Recruit bees for selected “e” sites according to pheromone levels at those sites (local search) to conduct searches in the neighborhood of the selected e sites and choose a bee with the highest fitness for each site The number of bees given by nb(Sj,t) recruited for a site Sj of e sites at time t is calculated from Equation (2.32) In this research, equations for nb(Sj,t), fs(Sj) và ph(Sj,t) calculations have been developed and improved in comparision with Equations of Dr Packianather (Cardiff University, United Kingdom) fs( S j )  fitness ( S j )  fitness (Se1 ) e   fitness(S )  fitness(S ) i 1 i e 1 10 (2.30) nb ( S j , t )  ph( S j , t  1) f s ( S j )  m.e (2.32) ph(S j , t )  ph(S j , t  1)   f s (S j ) nb (S j , t ) (2.34) e   ph(Si , t  1) f s (Si )    i 1 Where : fs(Sj) is the fitness score of site Sj Note that the fitness score fs(Sj) is normalized to smooth noise and suppress systematic variations m is the average number of bees at each site of e Se+1 is the best performing site among the non-selected sites ph(Sj,t) is pheromone value at site Sj The parameters α and β control the influence of the amount of pheromone that was available at site Sj from the previous iteration, and of the fitness score of site Sj on the bee recruitment The parameter ρ controls the evaporation or decay of pheromone  Step 6: Assign the remaining (n-e) bees to search randomly around the search space for new potential solutions These searches are carried out to avoid a local optimization  Step 7: At the end of the local and global search, the best bees from all sites are sorted according to their fitness  Step 8: Update new population  Step 9: Update pheromone level on each site by using Eq (2.34)  Step 10: : Stop the program if the convergence criteria is satisfied, otherwise go back to step The convergence criterion is that the fitness of the best bee of the new population is not improved after the specific number of iterations or after N iterations 11 Chapter EXPERIMENTAL SIMULATIONS OF KEY COMPONENT SIZE AND CONTROL STRATEGY PARAMETER OPTIMIZATION FOR HEV POWERTRAIN SYSTEMS EXPERIMENTAL SIMULATION OBJECTIVES Contents of experimental simulation in this dissertation are to get these following objectives : (1) Solving the problem of HEV powertrain system optimization by using BAs (2) Comparing the HEV powertrain system optimization quality between a basic bees algorithm and a pheromone-based bees algorithm (3) Comparing HEV powertrain system optimization results according to Vietnamses driving cycle and FTP driving cycle (4) Evaluating the reliability of HEV powertrain system optimization method using BAs EXPERIMENTAL SIMULATION CONTENTS Or Number List of Experimental Simulation Items Experimetal Simulations for Honda Insight 2000 Optimization of only control strategy parameters by using BBA according to Vietnamese driving cycle (CECDC) Optimization of only control strategy parameters by using BBA according to FTP driving cycle Optimization of only control strategy parameters by using PBA according to CECDC driving cycle Optimization of only control strategy parameters by using PBA according to 12 FTP driving cycle Simultaneous optimization of energy source sizes and control strategy parameters by using BBA according to CECDC driving cycle Simultaneous optimization of energy source sizes and control strategy parameters by using BBA according to FTP driving cycle Simultaneous optimization of energy source sizes and control strategy parameters by using PBA according to CECDC driving cycle Simultaneous optimization of energy source sizes and control strategy parameters by using PBA according to FTP driving cycle Experimetal Simulations for Toyota Prius 1998 Optimization of only control strategy parameters by using PBA according to CECDC driving cycle with different weighting factors of wi Simultaneous optimization of energy source sizes and control strategy 10 parameters by using PBA according to CECDC driving cycle with different weighting factors of wi Honda Insight 2000 The optimization objectives for Honda Insight 2000 with Vietnamese dynamic performance constraints as follows :  Min  G ( X )   Min( FC )  vmax  120km / h     11% t  200 m  20, s As optimization of only control strategy parameters using BBA and PBA (Tab 3-7; 3-8; 3-9 and 3-10 in Appendix) - BBA and PBA have determined a new value set of control strategy parameters as shown at columns (1), (2) and (3) in comparision with a current value set of control 13 strategy parameters of Honda Insight 2000 at column (0) according to CECDC and FTP driving cycles Comparing between the values of FC, vmax, min and t200m found by BBA, and PBA and the current value of FC, vmax, min and t200m of Honda insight 2000, for example, the value at column (1) and column (0) of Tab 3-7; 3-8; 3-9 and 3-10 show that FC is improved, and the maximum velocity, vmax, the grability, min, and the acceleration time, t200, are equivalent to the current value of Honda Insight 2000 - The power source sizes are kept in constant values when only optimizing control strategy parameters according to CECDC and FTP driving cycles Since the current total maximum power of ICE and EM of Honda Insight 2000 is greater than the required power so that Honda Insight 2000 can fulfill three Vietnamese dynamic constraint conditions So, the penalty function (2.24) and (2.28) not bring into play too much effectiveness All found optimization results show that min is greater than the required value of 11%, vmax is always greater than 120 km/h and t200 is always less than 20,4s - With different driving cycles, the value of optimal control strategy parameters are also different When using the same control strategy parameters for both FTP and CECDC will give different FC values - According to Tab 3-7; 3-8; 3-9 and 3-10, the optimization results of FC, vmax, min and t200m by using PBA are equivalent to using BBA, however, the convergence speed of PBA is higher than that of BBA about 20% As simultaneous optimization of both power source sizes and control strategy parameters using BBA and PBA (Tab 3-11; 3-12; 3-13 and 3-14 in Appendix) - BBA and PBA have determined a set of key component sizes of powertrain system represented by scaling factors, such as fc_trq_scale, mc_trq_scale and ess_cap_ scale, and new control strategy parameters from column (1) to column (5) in Tab 3-11; 3-12; 3-13 and 3-14 in comparision with the current value of power source sizes and control strategy parameters of Honda Insight 2000 at column (0) - Since required dynamic performances (vmax, min, t200m) are lower than current performances of Honda Insight 2000, the nescessary maximum power after 14 optimization is reduced For example, with power source sizes at column (1) in Tab 3-11, the maximum power of ICE is equal to 45%, the maximum power of EM is equal to 30,5% of their current power on Honda Insight 2000, the battery capacity is equal to 97% of current battery capacity - The penalty functions (2.24) and (2.28) built by author have helped optimization results of vmax, min and t200m to fufill Vietnamese dynamic constraint requirements, and also get minimum value of FC - Hybrid factor (HF) of current HEVs in market is often from 0,1 to 0,5 HFs of sets of results found in Tab 3-11; 3-12; 3-13 and 3-14 by using BBA and PBA are the same as HF of Honda Insight 2000 - With different driving cycles, the values of power souce sizes and optimal control strategy parameters are also different - Comparing the value of FC, vmax, min and t200m in Tab 3-11 and 3-12, or FC, vmax, min and t200m in Tab 3-13 and 3-14, PBA gives the same values of FC, vmax, min and t200m BBA as BBA However, the convergence speed of PBA is higher than that of BBA about 23% Toyota Prius 1998 The optimization objectives for Toyota Prius 1998 with Vietnamese dynamic performance constraints as follows :  Min  G ( X )   Min( w1.FC  w2 HC  w3 CO  w4 NO)   w 1 i  i 1  vmax  120km / h    11%  t200 m  20,5s   Where FC , HC , CO and NO are normalized 15 As optimization of only control strategy parameters using PBA (Tab 3-15 in Appendix) (1) The weighting factor wi effects too much to optimization results as follows: - The minimum found value of FC is inversely proportional to w1 as shown in Fig 3-6 When w1 = 0,85, the value of FC found by using PBA reduces 0,858 liter/km (12,5%) However, when w1 = 0,1, the minimum found value of FC is greater than 0,171 liter/100km in comparision with the current FC value of Toyota Prius 1998 Fig 3-6 Relation between w1 and FC - Due to  w  , when w1 is increased to reduce FC, weighting factors w2, w3 i 1 i and w4 have to be reduced The content of HC changes similarly to FC; HC is inversely proportional to w1 and proportional to w2 as shown in Fig 3-7 When w1 = 0,85 and w2 = 0,05, the content of HC reduces 0,049 g/km (4,9%) However, when w1 = 0,1 and w2 = 0,3 the minimum content of HC is greater than 0,065 g/km (6,5%) in comparision with the current HC content of Toyota Prius 1998 16 Fig 3-7a Relation between w1 and HC Fig 3-7b Relation between w2 and HC - The content of CO (g/km) is proportional to w1 and inversely proportional to w3 as represented in Fig 3-8 When w1 = 0,85 and w3 = 0,05, the content of CO found by PBA is greater 0,008 g/km (0,7%) than the current CO content of Toyota Prius 1998 However, when w1 = 0,1 and w2 = 0,3, the minimum content of CO is reduced 0,105 g/km (9,3%) 17 Fig 3-8a Relation between w1 and CO Fig 3-8b Relation between w3 and CO - The content of NO (g/km) is proportional to w1 and inversely proportional to w4, as shown in Fig 3-9 When w1 = 0,85 and w4 = 0,05, the content of NO found by PBA is greater 0,012 g/km (5,7%) than the current NO content of Toyota Prius 1998 However, when w1 = 0,1 and w4 = 0,3, the minimum NO content is reduced 0,07 g/km (33,2%) in comparision with the current NO content of Toyota Prius 1998 18 Fig 3-9a Relation between w1 and NO Fig 3-9b Relation between w4 and NO (2) The power source sizes are kept in constant values when only optimizing control strategy parameters according to CECDC driving cycle Since current total maximum power of ICE and EM of Toyota Prius 1998 is greater than the required power so that Toyota Prius 1998 can fulfill three Vietnamese dynamic constraint conditions So, the penalty function (2.28) not bring into play considerable effectiveness All found optimization results show that min is greater than the required value of 11%, vmax is always greater than 120 km/h and t200 is always less than 20,5s As simultaneous optimization of both power source sizes and control strategy parameters using PBA (Tab 3-19 in Appendix) (1) Comparing with optimization of only control strategy parameters, when simultaneous optimization of both power source sizes and control strategy parameters using PBA for Toyota Prius 1998, the weighting factors wi not effect considerably 19