Printer Friendly

A Predictive Energy Management Strategy Using a Rule-Based Mode Switch for Internal Combustion Engine (ICE) Vehicles.


The demand for electric power consistently increases as the number of electric components in the ICE vehicles grows [1]. In the ICE vehicles, an alternator is mechanically attached to the engine and rotates with a crankshaft by a pulley. The alternator generates electric energy through rotational motion of the crankshaft. In this process, the operation of the alternator affects the operating point of the engine [2]. In effect, this causes an increase in fuel consumption when supplying the electric energy for electric power systems. Therefore, research on an EM system that can limit the associated fuel consumption is necessary.

In order to improve the fuel economy, residual kinetic energy is used in this paper. The residual energy occurs when a vehicle decelerates and is usually consumed by brake force or friction. The wasted energy can be recycled to generate electric energy. Several systems have been developed to exploit residual energy, such as the kinetic energy recovery system (KERS) [3], [4] and hybrid electric vehicles (HEVs). For example, in a Formula One racing car, the Flywheel KERS has been applied [3]. In this system, a flywheel stores the kinetic energy as rotational energy from a driveshaft. For reuse of the kinetic energy, the KERS generally uses additional components as a low-friction flywheel. In the case of HEVs, a regenerative brake system has been installed [5], [6]. It converts the heat energy of brakes into electric power. The additional devices are also installed to the HEVs, as an electric motor is equipped for conversion and recovery of the electric energy.

This paper presents an EM strategy which can improve fuel economy without additional equipment for ICE vehicles. The EM strategy switches the control mode of the alternator according to predefined rules. The rules are based on the future duration time of the residual engine power. Electric energy production is varied along the control mode to reduce fuel consumption.

The proposed EM strategy progresses in two main stages. In the first stage, the future vehicle speed is predicted based on the Markov chain model that is derived from the driving database. The predicted speed is then used to estimate the future duration time of the residual engine power. In the second stage, an alternator control mode switch is operated based on the results of the first stage. There are three main modes for the switch: normal charge mode, high-rate charge mode, and low-rate charge mode. In the normal charge mode, the alternator operates in the same manner as it would in a conventional system. When there is residual engine power, the mode is switched to the high-rate charge mode. In this mode, the amount of electric power generation increases. Lastly, the low-rate charge mode is activated when the generation of residual energy is estimated in the near future.

In this mode, the alternator supplies minimum operating capacity. This strategy allows for the effective usage of residual power, and as a result, it reduces additional fuel consumption.

The remainder of this paper is organized as follows: Section II gives an overview of the EM system along with its system architecture. Then, Section III presents the stochastic prediction model for vehicle driving information. In Section IV, an EM using a rule-based mode switch is explained. In Section V, simulation results are presented. Finally, in Section VI, the conclusion is given.


System Overview

The proposed EM strategy is applied to the electric system of conventional ICE vehicles. In the target electric system, the alternator is controlled by a negative feedback controller using SOC value as a reference input. The proposed strategy provides a variable SOC setpoint to the feedback controller of the alternator.

As shown in Fig. 1, this EM consists of two sub-systems: a stochastic prediction system and a control mode switch. First, the prediction system estimates future duration time of the residual engine power based on input of current vehicle measurements. The prediction is based on a Markov chain model that is one of the representative stochastic models. Then, the control mode of the alternator is switched based on predefined switching rules. The rule determines the control mode according to the predicted information and the SOC of the battery. As the control mode switches, the set-point of the SOC changes as well. Finally, the error between the SOC set-point and current SOC values are used by the feedback controller of the alternator.


Relationship between Vehicle Speed and Residual Kinetic Energy

As mentioned in the previous section, the prediction system aims to estimate the future duration time of residual kinetic energy. However, it is difficult to predict the residual energy directly. Therefore, we used vehicle speed to predict the occurrence of residual energy. Correlation between the vehicle speed and the residual kinetic energy is shown in Fig. 2. It is observed that the residual power occurs when the vehicle decelerates. However, the deceleration does not always result in the residual power as shown in Fig. 2(c). Therefore, we applied a threshold for the magnitude of vehicle deceleration to distinguish the driver's demand. The threshold value was determined empirically. Consequently, the residual power is indicated by the value of vehicle acceleration under the threshold.

Markov Chain Model

In real driving, the prediction of vehicle speed is not deterministic, since the behavior of a vehicle is determined by the driver as well as various driving environment factors [7]. To model a non-deterministic system, a stochastic approach is used. In this paper, a Markov chain model is applied for prediction. This Markov chain model is based on an interval encoding method and multivariate Markov chain [8], [9].

We defined the state space X and partitioned it into a finite set with equal intervals. Each interval is assigned to Markov state [[bar.x].sub.j], which is the midpoint of the interval. With [[bar.x].sub.j] and predefined transition probabilities [[pi].sub.ij], a deterministic next state [x.sup.+] is predicted as

[x.sup.+] = [[summation].sup.M.sub.j=1][[pi].sub.ij][[bar.x].sub.j] (1)

Based on this interval encoding, the continuous real-valued state can be calculated [4]. In order to improve predictive accuracy, we selected three vehicle measurements as the Markov states: vehicle speed, vehicle acceleration, and engine speed.

Actual driving data were collected to design the Markov chain model. The data collection was repeated 32 times on the route of Fig. 3 in Seoul, Korea. The data was measured under similar conditions; no traffic jams were encountered, and there was a low volume of traffic. Fig. 4 shows the 32 velocity trajectories of the vehicle measured. 31 of the trajectories were used as training data for the Markov chain model, and the last one was used for verification simulation.

Evaluation Results

The Markov chain model of vehicle speed was evaluated through the simulation. The maximum prediction horizon (PH) of the Markov model is 10 seconds. PH indicates how long the prediction model can look ahead into the future. Fig. 5 shows the results in time domain. Each graph figures out different performances for each PH. As the PH increases, the accuracy of the prediction decreases. Quantitative analysis results are given in Table 1.


In the switching strategy of the alternator control mode, three modes are defined:

1. Mode 1 = normal charge mode (NC)

2. Mode 2 = high-rate charge mode (HC)

3. Mode 3 = low-rate charge mode (LC).

The control mode of the alternator is determined according to predefined rules, which is expressed by the "if-else" language as follows:

If [a.sub.veh](t) < [[theta].sub.a]

Mode(t + 1) = HC mode

Elseif [a.sub.veh](t) > [[theta].sub.a] and [SOC.sub.ResPwr] > [SOC.sub.Err]

Mode(t + 1) = LC mode


Mode(t + 1) = NC mode

At first, this EM basically runs in the NC mode. In the NC mode, the alternator operates identically to that of the conventional vehicles. It tries to maintain the SOC value in the stable region. The normal operating range for the batteries is usually 50-80% [10]. Therefore, the basic SOC set-point for NC mode is set to 70%.

At the next stage, the residual kinetic energy is confirmed through the current acceleration ([a.sub.veh]). If [a.sub.veh] is lower than the threshold ([[theta].sub.a]), it implies the occurrence of residual power, and the mode is shifted to HC mode. In the HC mode, the SOC set-point is set to high value and electric power production is increased. In this condition, the electric energy generation causes minimum fuel consumption by using the portion of the residual power.

Then, when [a.sub.veh] is higher than [[theta].sub.a], the amount of SOC charged by the HC mode in near future ([SOC.sub.ResPwr]) is estimated. In other words, the [SOC.sub.ResPwr] is calculated based on the duration time of residual power, which is estimated through the prediction system. Since the alternator generates maximum electric power with the residual power, [SOC.sub.ResPwr] is calculated as

[SOC.sub.ResPwr] = [S[??]C.sub.max] x [t.sub.ResPwr], (2)

where [S[??]C.sub.max] is maximum charge rate of SOC and [t.sub.ResPwr] is the predicted duration time of residual power. The [S[??]C.sub.max] is a coefficient which is dependent on specifications of the alternator and battery capacity. When [SOC.sub.ResPwr] > [SOC.sub.Err], it is confirmed that the residual energy recovery can supplement the lack of current SOC in the near future. [SOC.sub.Err] is calculated as

[SOC.sub.Err] = [SOC.sub.set] - [SOC.sub.meas], (3)

where [SOC.sub.set] is the normal SOC set-point and [SOC.sub.meas] is current measured SOC. In the LC mode, the SOC set-point is changed to low value. In this case, the electric energy generation is minimized to prevent unnecessary fuel consumption.

Finally, when there is no chance for residual energy recovery, the EM keeps the NC.


Simulation Environment

The proposed EM strategy was validated through simulation using the Autonomie which was developed by Argonne National Laboratory (ANL) [11]. Specifications of the vehicle model are as in the following table.

For evaluation, the proposed strategy--that is, the CMS--was compared to a proportional control strategy whose SOC set-point is constant. The proportional control strategy is provided by the Autonomie alternator control model.

The test driving cycle is one of the collected velocity trajectories in the Fig. 4, as mentioned in Section III. Total time and distance of the cycle are 238 seconds and 17.11 km respectively. The other analysis results of the test driving cycle are shown in the Fig. 6.

Simulation Results

Simulation results show that the proposed EM can improve fuel economy. Fig. 7 presents trajectories of normalized alternator command and SOC according to change of control mode. The alternator command is maximized when the residual power occur (Residual power indicator = 1). The alternator is turned off by the CMS when the residual kinetic energy is detected in future. On the other hand, it operates for the entire duration of time in results of the proportional control strategy. Both of the two strategies keep SOC values in the stable range.

As a result, Fig. 8 and Table 3 show that total fuel consumption is reduced by 0.48%. In terms of electric energy production, the fuel consumed by the alternator is reduced by 42.86%.


Advanced energy management strategies are developed to efficiently supply electric energy while reducing fuel consumption. In this paper, the energy management strategy based on the mode switch of the alternator control has been proposed. The alternator control mode is switched according to predefined rules. The rules are based on the residual kinetic energy for energy recovery. In order to estimate the duration time of residual energy, this strategy uses future vehicle speed predicted by a Markov chain model. The proposed strategy is evaluated through simulation and compared with the conventional strategy. The total fuel consumption is reduced by 0.48%. In the case of the fuel consumed based on the operation of the alternator, the proposed EM achieves fuel saving by 42.86%. This method is suitable for implementation in the conventional vehicle because it does not require the additional hardware. There is also the potential for expansion in HEVs.

The future work will focus on application of intelligent traffic system (ITS). ITS information such as traffic signs and traffic lights can improve performance of the stochastic prediction system.


[1.] Kassakian, J. G. Miller J. M., and Traub, N. "Automotive electronics power up," IEEE Spectrum. 37(5):34-39, 2000, doi:10.1109/6.842132

[2.] Koot, M., Kessels J.T.B.A., Jager B.d., Heemels W.P.M.H., , "Energy management strategies for vehicular electric power systems," IEEE Transactions on Vehicular Technology. 54(3):771-782, 2005, doi:10.1109/TVT.2005.847211.

[3.] Gabriel-Buenaventura, A. and Azzopardi B., "Energy recovery systems for retrofitting in internal combustion engine vehicles: A review of techniques," Renewable and Sustainable Energy Reviews. 41:955-964, 2015.

[4.] Gramling, J., "Fully Integrated IVT-Regenerative Braking," SAE Technical Paper 2014-01-1727, 2014, doi:10.4271/2014-01-1727.

[5.] Nakamura, E., Soga, M., Sakai, A., Otomo, A., "Development of Electronically Controlled Brake System for Hybrid Vehicle," SAE Technical Paper 2002-01-0300, 2002, doi:10.4271/2002-01-0300.

[6.] Brandenburg, Larry R., and King Edward T.. "Hybrid electric vehicle regenerative braking energy recovery system." U.S. Patent No. 5, 291,960, 1994.

[7.] Karbowski, D., Kim N., and Rousseau A., "Route-Based Online Energy Management of a PHEV and Sensitivity to Trip Prediction," 2014 IEEE Vehicle Power and Propulsion Conference (VPPC), doi:10.1109/VPPC.2014.7007126.

[8.] Filev, D.P. and Kolmanovsky I., "A generalized Markov Chain modeling approach for on board applications," The 2010 International Joint Conference on Neural Networks (IJCNN), doi:10.1109/IJCNN.2010.5596713.

[9.] Filev, D.P. and Kolmanovsky I., "Generalized Markov Models for Real-Time Modeling of Continuous Systems," IEEE Transactions on Fuzzy Systems. 22(4):983-998, 2014, doi:10.1109/TFUZZ.2013.2279535.

[10.] Amphlett, John C., "Dynamic interaction of a proton exchange membrane fuel cell and a lead-acid battery." Journal of power Sources 65(1):173-178, 1997, doi:10.1016/S0378-7753(97)02472-5.

[11.] Moawad, A., Kim, N., and Rousseau, A., "Impact of Technology on Electric Drive Fuel Consumption and Cost," SAE Technical Paper 2012-01-1011, 2012, doi:10.4271/2012-01-1011.


This research was supported by "Practical Development of Wireless Low-Floor Tram"(15RTRP-B067379-03) project funded by Korean Ministry of Land, Infrastructure and Transport, by the Industrial Strategic Technology Development Program "Development of Next Generation E/E Architecture and Body Domain Unit for Automotive Body Domain" (10060068) funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea), by Energy Resource R&D program(2006ETR11P091C) under the Ministry of Knowledge Economy, Republic of Korea, by the Industrial Strategy Technology Development Program of Ministry of Trade, Industry and Energy (No. 10042633), and by the BK21 plus program (22A20130000045) under the Ministry of Education, Republic of Korea.


ICE - Internal combustion engine

HEV - Hybrid electric vehicle

SOC - State of charge

CMS - Control mode switch

ITS - Intelligent traffic system

EM - Energy management

Haksu Kim, Jaewook Shin, and Myoungho Sunwoo

Hanyang University

Table 1. RMSE and R-squared values for each PH

PH [s]     RMSE [km/h]     R-squared

 1          1.68            0.9998
 4          4.72            0.9861
 7          7.99            0.9686
10         10.29            0.9452

*RMSE: root mean square error

Table 2. The specification of the target vehicle model

Component       Specification

Engine          110kW 2.2L Spark Ignition Direct Injection (SIDI)
Transmission    Automatic transmission (AT)
Alternator      2kW
Battery         12V 66Ah lead-acid battery
Electric load   200W

Table 3. Comparison of the two strategies in terms of fuel consumption

Control strategy         Fuel consumption [g]
                         Total                  By alternator

Constant SOC set-point   264.6                   2.80
CMS                      263.4                   1.60
Saving                     0.48%                42.86%
COPYRIGHT 2017 SAE International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Kim, Haksu; Shin, Jaewook; Sunwoo, Myoungho
Publication:SAE International Journal of Engines
Date:Apr 1, 2017
Previous Article:Modeling Non-Premixed Combustion Using Tabulated Kinetics and Different Fame Structure Assumptions.
Next Article:Individual Cylinder Air-Fuel Ratio Control for Engines with Unevenly Spaced Firing Order.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters