Real‐time energy management for commute HEVs using modified A‐ECMS with traffic information recognition
2018; Institution of Engineering and Technology; Volume: 13; Issue: 4 Linguagem: Inglês
10.1049/iet-its.2018.5274
ISSN1751-9578
Autores Tópico(s)Vehicle emissions and performance
ResumoIET Intelligent Transport SystemsVolume 13, Issue 4 p. 729-737 Research ArticleFree Access Real-time energy management for commute HEVs using modified A-ECMS with traffic information recognition Yang Li, Yang Li School of Electrical Engineering, Yanshan University, Qinhuangdao, People's Republic of ChinaSearch for more papers by this authorXiaohong Jiao, Corresponding Author Xiaohong Jiao jiaoxh@ysu.edu.cn orcid.org/0000-0001-7276-6062 School of Electrical Engineering, Yanshan University, Qinhuangdao, People's Republic of ChinaSearch for more papers by this author Yang Li, Yang Li School of Electrical Engineering, Yanshan University, Qinhuangdao, People's Republic of ChinaSearch for more papers by this authorXiaohong Jiao, Corresponding Author Xiaohong Jiao jiaoxh@ysu.edu.cn orcid.org/0000-0001-7276-6062 School of Electrical Engineering, Yanshan University, Qinhuangdao, People's Republic of ChinaSearch for more papers by this author First published: 17 January 2019 https://doi.org/10.1049/iet-its.2018.5274Citations: 3AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract To further improve fuel consumption performance of hybrid electric vehicles (HEVs) running on commute route in the face of time-varying traffic information, this paper investigates a real-time energy management strategy based on the adaptive equivalent consumption minimization strategy (A-ECMS) framework with traffic information recognition. The proposed management strategy integrates the global near optimization and the real-time performance. The simple traffic recognition is constructed by utilising k-means clustering algorithm to deal with the historical traffic data to form four clusters. The adaptive equivalence factor of the A-ECMS is designed as a three-dimensional mapping on each cluster and the system states by employing stochastic dynamic programming (SDP) policy iteration to solve offline the stochastic optimal control problem formulated by each cluster statistical characteristic. In real-time energy management controller online, the instantaneous power split is performed by the ECMS with a proper equivalent factor, which is obtained from mappings according to the cluster recognised by the current traffic situation and the state-of-charge (SOC). The effectiveness of the designed control strategy is verified by the simulation test conducted on GT-suite HEV simulator over real driving cycles. 1 Introduction Unlike traditional fossil-fuelled vehicles, hybrid electric vehicles (HEVs) can use the electrical power stored in the battery to propel the vehicle no matter non-plug-in or plug-in type (the battery can or not be charged through an external power grid). Thus, energy management strategies (EMS) controlling the engine and electrical power flows are crucial to reducing fuel consumption for HEVs. The objective of the EMS is as much as possible to reduce the fuel consumption under the condition of satisfying the overall vehicle power demands and the battery state-of-charge (SOC) remaining within a prescribed range (non-plug-in) or achieving the minimum limit value (plug-in) at the end of driving path. As a result, the dynamic programming (DP) [1, 2] based on global optimization would be likely the ideal design for the EMS. Unfortunately, the driving cycle is unknown in advance, which results in the infeasibility of the EMS designed offline based on DP in practice. Consequently, finding a strategy that can cope with the uncertain driving cycle in practice and whose effects is close to the control performance of the DP in theory has recently drawn increased attention in the energy management problem of HEVs. To this end, stochastic DP (SDP) utilising the probability distribution from multiple historic driving cycles is naturally suggested as possible alternative [3, 4]. In fact, the essential of SDP is to predict the most likely future driving cycle by utilising the statistical characteristics extracted from a mass of traffic information data, and then to achieve the near-optimal energy management results. Meanwhile, model predictive control (MPC) with the finite receding horizon optimization is to predict the short-term driving cycle in the near future and then to execute the optimal solution over this short predicted horizon [5-7]. Inspired by this prediction, the other alternative to cope with the uncertain driving cycle in practice is the driving pattern recognition (DPR) technique. Various DPR methods [8-11] are employed to classify the typical driving patterns from historical driving information and then to recognise the current driving pattern so as to guarantee the effects of the optimal management results. From the view of real-time management, the equivalent consumption minimization strategy (ECMS) based on an instantaneous optimization [12] is popular, which is designed by minimising the instantaneous cost function consisting of the fuel consumption and an equivalent fuel consumption related to the battery SOC variation. The definition of such a cost function requires an equivalence factor for comparing the electrical energy with the fuel energy. It has shown that ECMS can be regarded as a realization of Pontryagin's minimum principle (PMP)-based global optimization problem in the whole driving cycle when the chosen equivalence factor in ECMS is link to the optimal costate of PMP [13-15]. Therefore, the equivalence factor in ECMS should be varied with the different driving pattern considering the overall power conversion efficiencies and battery SOC constraint. Accordingly, various modified ECMS approaches, adaptive equivalent consumption minimization strategies (A-ECMS), have been proposed [16-21]. Based on the above analysis, for both globally suboptimal and implementable energy management, this paper provides a novel real-time energy management strategy based on A-ECMS with traffic information recognition for commuter HEVs. The contributions of this paper are there aspects. One is k-means algorithm is utilised to deal with the traffic information data resulting from the collected historical commuter driving cycles into several different clusterings. The obtained cluster centres can be exploited to serve as a real-time traffic information recognitor. Meanwhile, the statistical characteristics in every clustering traffic information is captured as a stochastic model with average acceleration as random disturbance. The second is finding adaptive equivalence factor of A-ECMS is transferred into solving stochastic optimal problem in each clustering by policy iteration algorithm. The adaptive equivalence factor is built as a three-dimensional mapping on each cluster and system state. The third is in real-time energy management controller online, the instantaneous power split is performed by the ECMS with a proper equivalent factor that is obtained from mappings according to the cluster recognised by the current traffic situation, velocity, and SOC. Thus, the difference with the existing results has two points. One is that both the global near optimization and the real-time performance of the EMS in practice are guaranteed by the ECMS with the quickly obtained equivalence factor from the mapping on the current driving pattern and the current system states. The other is the mapping of the equivalence factor has better adaptivity to the time-varying traffic information because it is obtained by utilising the policy iteration algorithm of infinite-horizon SDP and the driving pattern recognizor. The remainder of this paper is arranged as follows. In Section 2, the traffic information model is introduced. In Section 3, A-ECMS with traffic information recognition is presented and applied. In Section 4, the validation results carried on the GT-suite simulator are illustrated. Conclusion is given in Section 5. 2 Traffic information modelling Here, in order to obtain the adaptive equivalence factor of the A-ECMS closely related to traffic information and system states of a HEV, a kind of traffic information model is established by dealing with the collected historical driving cycles data and employing the k-means cluster algorithm. Details are presented as follows. First, three weeks driving cycles' data of a non-plug in HEV on a commuter route provided by [22] can be exploited to capture the characteristics of the traffic condition. Meanwhile, in order to guarantee the sufficient driving cycles' data for the statistic characteristics, the multiple trips are generated by utilising the scaling method [4]. Then, the k-means algorithm is utilised to cluster these velocity profiles according to some feature parameters reflecting traffic conditions. Here, the 15 feature parameters in a certain distance (here 200 m), listed in Table 1, are chosen to serve as the traffic information recognition. Where both average velocity and average acceleration are versus the certain distance, that is, , , n is the sample number in the 200 m distance and is the sample period. Table 1. Fifteen feature parameters for classification Notation Meaning Unit average velocity vs distance [km/h] percent of velocity of 0–10 km/h [%] percent of velocity of 10–20 km/h [%] percent of velocity of 20–30 km/h [%] percent of velocity of 30–40 km/h [%] percent of velocity of 40–50 km/h [%] percent of velocity of >50 km/h [%] average acceleration vs distance [] percent of time accelerating [%] percent of time decelerating [%] percent of time cruising [%] percent of time stopping [%] percent of distance acceleration [%] percent of distance deceleration [%] percent of distance cruising [%] The k-means algorithm [23] is one of the mostly used clustering algorithms, which can be as categorization tool. Clustering in N -dimensional Euclidean space is the process of dividing a given set into m groups based on a clustering criterion. The criterion can be formulated as: (1) where m presents the number of clusters centroid, and denotes the number of elements belonging to i th cluster. is the j th elements of the i th cluster and is the i th cluster centroid. The cluster centroid can be expressed as follows: (2) In general, four steps are involved in the k-means algorithm, which are briefly listed below: (i) Select initial cluster centres from the samples . (ii) Allocate the samples to the cluster centres , if , . (iii) Compute new cluster centres by (ii). (iv) If stop algorithm, else go to step (ii). The change to result of the k-means algorithm is to cluster velocity profiles into m = 4 types, as shown in Fig. 1. Moreover, the centroid of the four clusters are obtained as follows, which can be exploited to serve as a real-time traffic information recognitor. (3) Fig. 1Open in figure viewerPowerPoint Results of the k-means algorithm for velocity profiles From Fig. 1, it can be found that the cluster 1 includes very low velocity such as facing traffic jam or waiting for traffic lights, the cluster 2 presents relatively low driving speed conditions, the cluster 3 contains relatively medium speed and some traffic jam conditions and the cluster 4 has medium speed and smooth traffic conditions. Each cluster centroid obtained can be served as recognising traffic information in the implementation of the real-time energy management strategy. Finally, a stochastic model of each cluster is established according to the probability of the average acceleration. Under the assumption of a normal distribution, the mean and standard deviation of the normal cumulative function of each cluster can be obtained by calculating the tested data. The process of data processing is presented in Fig. 2. Fig. 2Open in figure viewerPowerPoint Process of the driving cycles data processing (a) collecting speed data, (b) classifying real driving cycles, (c) probability distribution of average acceleration Specifically, the speed data are collected through the 200 m equidistant sliding window and the 0.02 s sample period. According to k-means results, the 15 days real driving cycles are divided into four categories. The probability distribution of average acceleration of the four categories are calculated, shown in Fig. 2c. 3 A-ECMS with traffic information recognition Here, a non-plug-in commuter HEV described in JSAE-SICE Benchmark Problem 2 of [22] is considered as the plant for the design and implementation of the real-time EMS. However, this kind of A-ECMS presented is not limited to the particular HEV type and can be applied to other HEVs, and even PHEVs with the modification of the prescribed requirement for SOC in driving cycle. 3.1 Powertarin model for optimization design of A-ECSM The simplified architecture of the considered HEV is shown in Fig. 3. The major components are the engine, the two electric machines, a battery pack, and a planetary gear set. Vehicle Model: During driving process, the vehicle motion equation can be written as: (4) where M [kg] and v [m/s] are the weight and the velocity of vehicle, respectively, and g is the acceleration of gravity. [N.m] and [N.m] are the axle torque on the differential gear and the brake torque. is the transmission efficiency of differential gear, is coefficient of rolling resistance, is air density, A is frontal area of vehicle, is drag coefficient, and is road angle. Transmission Model: Under the assumption of no friction losses and rigid connection, the relations of torque and speed among the ICE, M/G1, and M/G2 can be expressed as: (5) Fig. 3Open in figure viewerPowerPoint Configuration of the non-plug-in HEV drive train The dynamics of the generator, engine and motor are, respectively, described as follows: (6) where T [N.m], [rad/s] and J [] denote torque, speed and inertia, respectively. The subscript g, m and e mean M/G1, M/G2 and ICE, respectively. are the radii of the ring gear and the sun gear. F is the internal force on pinion gears, is the final differential ratio. Battery Model: SOC dynamics of battery based on the equivalent-circuit model is described as: (7) where [V], [Ah], [W], and are open circuit voltage, maximum charge capacity, power, and internal resistance of battery, respectively. Energy Flow: The relationship of the energy flow of the powertrain can be described as: (8) (9) where are generator and motor efficiency. is the sign function, which larger than zero presents motoring state, otherwise generating state. is the propelling power demand. Consumption: The consumptions of ICE and the electrical power (battery) are defined as: (10) (11) where [g/s] is engine fuel consumption, [g/kWh] is brake-specific fuel consumption, and [g/s] denotes the converted equivalent consumption from the electrical power by applying the lower heating value of the fuel [J/g] and equivalence factor . 3.2 Design for adaptive equivalence factor of A-ECMS Based on the established statistic characteristics model of traffic information, finding appropriate adaptive equivalent factor of A-ECMS in each cluster can be converted into a stochastic non-linear and constrained optimal control problem. Where the battery SOC and vehicle velocity are the system states and the average acceleration of traffic flow is regarded as the stochastic disturbance. For the real-time performance of the designed EMS in practice, the authors focus on the formulated stochastic optimization problem can be easily solved to obtain a closed-loop optimal solution on the system states. That way the solved adaptive equivalence factor can be formed by a mapping on the system states and traffic information so that in the implementation of A-ECMS during real driving cycles the adaptive equivalence factor can be real-time adjusted according to the current system states and traffic conditions. Fortunately, formulating an infinite-horizon SDP with a discount factor <1 can achieve the purpose to derive a convergent optimal solution, a time-invariant casual state-dependent closed-loop control policy by the policy iteration algorithm. Therefore, in this subsection, detail is presented on obtaining adaptive equivalence factor by the discrete infinite horizon SDP and the policy iteration algorithm. First, obtaining adaptive equivalence factor in each cluster is transformed into the discrete infinite-horizon SDP problem defined as minimising the expected sum of a running cost function: (12) with a discount factor , and the running cost function: subject to the dynamic constraints for the system states: (13) with where is average velocity vs the distance , is average acceleration in the distance , k is the sample point according to the distance . The stochastic disturbance is and the control inputs are constrained to a given non-empty subset with . , are finite sets with , which are determined by the following physical constraints: (14) where , , are the minimum and maximum values of the speed, torque, and battery SOC under the system physical characteristics, respectively. Then, in order to obtain the suitable adaptive equivalence factor, this infinite-horizon SDP optimization problem is solved using the policy iteration algorithm with the following procedure: (i) Initialization step: setting the sets according to the constraints (14), and give a initial stationary policy , which does not affect the optimization result. (ii) Policy evaluation step: calculate the corresponding cost function based on matrixes of transition probabilities and cost per stage and with respect to (iii) Policy improvement step: compute a new policy by (iv) Judge the convergence condition , if satisfied go on; else return back to (ii). (v) Obtain the optimal control policy . By utilising the policy iteration algorithm for the SDP optimization problem in each cluster, the final control policy optimised offline is a static state-feedback type on SOC and vehicle velocity, which just is a 2D-mapping on SOC and vehicle velocity in each cluster. According to the four statistical probability characteristics of the clusters, the four mappings obtained are shown in Fig. 4. Fig. 4Open in figure viewerPowerPoint Maps of adaptive equivalent factors of A-ECMS (a) cluster 1, (b) cluster 2, (c) cluster 3, (d) cluster 4 3.3 Implementation of A-ECMS Once the cluster centres (3) and the adaptive equivalence factor mapping Fig. 4 are obtained, these mappings are embedded in the energy management control system of HEV in actual online operation. Specifically, the architecture of the A-ECMS with traffic information recognition is depicted as Fig. 5, which includes two parts: one part is the cluster centre and proper equivalence factor obtained offline by k-means algorithm and policy iteration algorithm, respectively, another is the designed energy management controller applied in actual operation. Fig. 5Open in figure viewerPowerPoint Architecture of A-ECMS with traffic information recognition As is shown in Fig. 5, in the first, real-time traffic situation of the latest driving certain distance is recognised by utilising the cluster centres (3) obtained by k-means clustering algorithm, that is, the current driving pattern is classified to one of the four clusters. The adaptive equivalence factor is generated online by looking up the optimised equivalence factor mappings as Fig. 4 according to the external traffic situation, cluster type and average speed, and internal system factor SOC. Then, the selected proper equivalence factor can help make a smart decision at the each time step so that the A-ECMS with traffic information recognition can distribute the energy flow to propel the HEV as conventional ECMS minimising the instantaneous cost function, the equivalent fuel consumption, to optimise the control inputs, engine torque , and speed . Moreover, the new collected driving cycles can update and enrich the speed database, which can amend the equivalence factor map and cluster centres to improve the effective of the energy management controller. 4 Simulation verification on GT-suite test Here, the simulation results operating on different driving cycles in GT-suite virtual test platform are given to demonstrate the effectiveness of the proposed energy management strategy. The block diagram of the Toyota prius system with the proposed strategy is shown in Fig. 6. The basic parameters of GT-suite HEV model are listed in Table 2. Table 2. Basic vehicle parameters applied in the simulation Notation Meaning Value[Unit] M vehicle mass 1460[] air density 1.293[] air drag coefficient 0.33[-] A frontal area of vehicle 3.8[m2] coefficient of rolling resistance 0.015[-] final differential gear ratio 4.113[-] radius of the tire 0.2982[m] radius of the ring gear [m] radius of the sun gear [m] maximum continuous ICE power 51[kW] maximum continuous M/G2 power 25[kW] peak rating power of M/G2 50[kW] maximum continuous M/G1 power 15[kW] peak rating power of M/G1 30[kW] battery maximum charge capacity 6.5[Ah] Fig. 6Open in figure viewerPowerPoint Diagram of HEV with the proposed A-ECMS The real driving cycles of the first Monday, the second Thursday, and the second Tuesday are selected as speed profiles for the simulation. The simulation results are presented in Figs. 7-9, where from top down, there are the curves of referenced and actual speed profile, equivalence factor, SOC trajectory, fuel consumption, engine operating points, working mode (cluster type), engine-generator-motor torques, and engine-generator-motor speeds, respectively. Fig. 7Open in figure viewerPowerPoint Simulation for the real driving cycle on the 1st Monday Fig. 8Open in figure viewerPowerPoint Simulation for the real driving cycle on the 2nd Thursday Fig. 9Open in figure viewerPowerPoint Simulation for the real driving cycle on the 2nd Tuesday From Figs. 7-9, it can be seen that the vehicle speed trajectories are almost identical, the equivalence factor is time-varying with the driving pattern, the SOC is limited nearby reference SOC, and the operating points of the ICE are distributed highly surrounding the best operating line that means the ICE is working efficiently. The cluster type is determined by the latest driving speed information and the pattern 1-4 correspond to the very low and zero velocity, the relatively low driving speed, the relatively medium speed, and the medium and smooth speed, respectively. Furthermore, the engine torque in the pattern 1 and 2 is almost zero and the motor supplies the torque to vehicle, and in most of the patterns 3 and 4, the engine and motor together supply the torque to vehicle. Furthermore, the comparison results with the other strategies are given to illustrate the advantage of the proposed strategy. First, energy management (EX) [22] and switching control (SPSA) algorithms [24] are considered since the results shown in [22, 24] both are researched on the same benchmark testing environment and condition. The compared results operating on the driving cycles of the second Thursday and the first Wednesday are shown in Figs. 10 and 11, respectively. From Figs. 10 and 11, it is observed that the all controllers performs precisely in tracking the vehicle speeds to the target vehicle speeds, which means good drivability of all methods. Another observation is that the trajectories of SOC are maintained within its physical limits because regenerative braking power and the excess engine power can be transformed to the electrical form through the generator and then pumped into the battery. The most important observation is that the fuel consumption of the designed controller is minimal, and the ICE operation points of the different controllers can illustrate the respective engine working efficiency. Then, the comparisons with the DP and the conventional ECMS (equivalence factor is fixed) are also presented. Similarly, the simulation results of the three strategies operating on the driving cycles of the second Thursday and the first Wednesday are shown in Figs. 12 and 13, respectively. From Figs. 12 and 13, it is also observed that the all strategies have good drivability with precisely tracking the target vehicle speeds and can remain the SOC within the prescribed range (0.5–0.65). Further observed that the fuel consumption of the designed controller is much nearer to that of the DP than that of the conventional ECMS strategy. Fig. 10Open in figure viewerPowerPoint Comparison with EX and SPSA strategies for the driving cycle on the 2nd Thursday Fig. 11Open in figure viewerPowerPoint Comparison with EX and SPSA strategies for the driving cycle on the 1st Wednesday Fig. 12Open in figure viewerPowerPoint Comparison with DP and CECMS strategies for the driving cycle on the 2nd Thursday Fig. 13Open in figure viewerPowerPoint Comparison with DP and CECMS strategies for the driving cycle on the 1st Wednesday In addition, to facilitate analysis of the comparison results on the fuel consumption of the different strategies, the respective fuel efficiency expressed by the fuel consumption per hundred kilometres [L/100 km] resulting from Figs. 10-13 are summarised in Table 3, where the quantitative value is calculated by Petrol No.93 with the density 0.725 kg/L. This table indicates that the control effect of the proposed A-ECMS is much nearer to that of the DP than that of other strategy, the proposed A-ECMS can improve the fuel efficiency by , , and compared with the strategies in [22, 24] and the conventional ECMS, respectively. Table 3. Fuel consumption of different strategies [L/100 km] EX SPSA CECMS DP Designed wed1 5.015 4.527 5.615 3.566 3.990 thurs2 4.730 4.177 4.581 3.547 3.940 5 Conclusions Here, various factors, such as traffic information, vehicle velocity and battery SOC, are utilised to design a kind of A-ECMS with traffic information recognition for a commute HEV by k-means algorithm and SDP algorithm. Considering the HEV running on fixed route, the historical traffic information is analysed to obtain the clusters centroid of the traffic information and to extract the statistical characteristic of averaged acceleration in different traffic condition. Meanwhile, finding adaptive equivalence factor for A-ECMS in the implementation is converted into a discrete stochastic optimization problem in each cluster, and establishing equivalence factor mapping on the system states is solved by the policy iteration algorithm offline. Thereby, due to capability of finding suitable adaptive equivalence factor for different driving cycles in actual operation, the proposed A-ECMS can effectively improve the reduction of fuel consumption by utilising the proper equivalence factor managing the engine operating nearby optimal line, and render the battery SOC to remain within a prescribed range during driving path. The comparison results with the three other EMS and the benchmark test DP also show that fuel efficiency is improved using the proposed A-ECMS for different driving conditions and is nearer to that of the DP. 6 Acknowledgments This work was supported by the National Natural Science Foundation of China (Grant No. 61573304) and the Natural Science Foundation of Hebei Province (Grant No. F2017203210). 7 References 1Lin C.C. Peng H. Grizzle J.W.: 'Power management strategy for a parallel hybrid electric truck', IEEE Trans. Control Syst. Technol., 2003, 11, pp. 839– 849 2Liu J. Peng H.: 'Modeling and control of a power-split hybrid vehicle', IEEE Trans. Control Syst. Technol., 2008, 16, pp. 1242– 1251 3Jiao X. Shen T.: 'SDP policy iteration-based energy management strategy using traffic information for commuter hybrid electric vehicles', Energies, 2014, 7, pp. 4648– 4675 4Zeng X. Wang J.: ' Stochastic optimal control for hybrid electric vehicles running on fixed routes'. American Control Conf., Chicago, American, 1–3 July 2015 5Borhan H. Vahidi A. Phillips A.M. et al.: 'MPC-based energy management of a power-split hybrid electric vehicle', IEEE Trans. Control Syst. Technol., 2012, 20, pp. 593– 603 6Di Cairano S. Bernardini D. Bemporad A. et al.: 'Stochastic MPC with learning for driver-predictive vehicle control and its application to HEV energy management', IEEE Trans. Control Syst. Technol., 2014, 22, pp. 1018– 1031 7Zhang J. Shen T.: 'Real-time fuel economy optimization with nonlinear MPC for PHEVs', IEEE Trans. Control Syst. Technol., 2016, 24, pp. 2167– 2175 8Wang J. Huang Y. Xie H. et al.: ' Driving pattern recognition and energy management for extended range electric bus'. 2014 IEEE Vehicle Power and Propulsion Conf. (VPPC), Coimbra, Portugal, 27–30 October 2014 9Zhang S. Xiong R.: 'Adaptive energy management of a plug-in hybrid electric vehicle based on driving pattern recognition and dynamic programming', Appl. Energy, 2015, 155, pp. 68– 78 10Lei Z. Qin D. Liu Y. et al.: 'Dynamic energy management for a novel hybrid electric system based on driving pattern recognition', Appl. Math. Model., 2017, 45, pp. 940– 954 11Shen P. Zhao Z. Zhan X. et al.: 'Optimal energy management strategy for a plug-in hybrid electric commercial vehicle based on velocity prediction', Energy, 2018, 155, pp. 838– 852 12Sciarretta A. Guzzella L.: 'Control of hybrid electric vehicles', IEEE Control Syst. Mag., 2007, 27, pp. 60– 70 13Kim N. Cha S.W. Peng H.: 'Optimal control of hybrid electric vehicles based on Pontryagin's minimum principle', IEEE Trans. Control Syst. Technol., 2011, 19, pp. 1279– 1287 14Kim N. Cha S.W. Peng H.: 'Optimal equivalent fuel consumption for hybrid electric vehicles', IEEE Trans. Control Syst. Technol., 2012, 20, pp. 817– 825 15Ahmadizadeh P. Mashadi B. Lodaya D.: 'Energy management of a dual-mode powersplit powertrain based on the Pontryagin's minimum principle', IET Intell. Trans. Syst., 2017, 11, (9), pp. 561– 571 16Musardo C. Rizzoni G. Guezennec Y. et al.: 'A-ECMS: an adaptive algorithm for hybrid electric vehicle energy management', Eur. J. Control, 2005, 11, pp. 509– 524 17Sivertsson M. Sundstrom C. Eriksson L.: 'Adaptive control of a hybrid powertrain with map-based ECMS', IFAC Proceedings Volumes, 2012, 44, pp. 2949– 2954 18Chen H. Kessels J.T.B.A. Weiland S.: ' Adaptive ECMS: a causal set-theoretic method for equivalence factor estimation'. 4th IFAC Workshop on Engine and Powertrain Control, Simulation and Modeling, Ohio, USA, 6–8 August 2015 19Sun C. Sun F. He H.: 'Investigating adaptive-ECMS with velocity forecast ability for hybrid electric vehicles', Appl. Energy, 2017, 185, pp. 1644– 1653 20Yang C. Du S. Li L. et al.: 'Adaptive real-time optimal energy management strategy based on equivalent factors optimization for plug-in hybrid electric vehicle', Appl. Energy, 2017, 203, pp. 883– 896 21Yang S. Wang W. Zhang F. et al.: 'Driving style-oriented adaptive equivalent consumption minimization strategies for HEVs', IEEE Trans. Vehicular Technol., 2018, 67, pp. 9249– 9261, doi: 10.1109/TVT.2018.2855146 22Yasui Y.: ' JSAE-SICE benchmark problem 2: fuel consumption optimization of commuter vehicle using hybrid powertrain'. Proc. 10th World Congress on Intelligent Control Automation, Beijing, China, 6–8 July 2012 23Pena J.M. Lozano J.A. Larranaga P.: 'An empirical comparison of four initialization methods for the k-means algorithm', Pattern Recognit. Lett., 1999, 20, pp. 1027– 1040 24Ahmad M.A. Azuma S.I. Baba I. et al.: 'Switching controller design for hybrid electric vehicles', SICE J. Control, 2014, 7, pp. 273– 282 Citing Literature Volume13, Issue4April 2019Pages 729-737 FiguresReferencesRelatedInformation
Referência(s)