Transportation Technology R&D Center Route-Based Energy Management for PHEVs: A Simulation Framework for Large-Scale Evaluation Dominik Karbowski, Namwook Kim, Aymeric Rousseau Argonne National Laboratory, USA The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory ("Argonne"). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract. DE-AC2-6CH11357. The U.S. Government retains for itself, and others acting on its behalf, a paid-up nonexclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, by or on behalf of the Government.
Optimal Energy Management of xevs Needs Trip Prediction Vehicle energy use can be reduced through application of control theory or fine tuning: Dynamic Programming (DP): finds the global optimum for the command law Instantaneous optimization: o ECMS: Equivalent Minimization Consumption Strategy o PMP: Pontryagin Minimization Principle All techniques require knowledge of the trip! Increased connectivity and increased availability of data opens the door to trip prediction GPS Sensors Geographical Information mph 6 4 2 1 2 3 Miles Trip prediction Live Traffic Situation On-board Computing Cloud Computing 2
Our Vision for Route-Based Control GPS Scope of Argonne s Research Driver s Input OR Current Position Destination Itinerary Computation Detailed Segment-by- Segment Information mph Route Prediction 6 4 2 Speed & Grade Route-based Optimization Optimal Control Optimal Energy Mgmt Pattern Recognition Live Traffic Average traffic speed 1 2 3 Miles Original research on speed prediction, an often overlooked problem Research on implementable solutions for routebased control Evaluation of real-world benefits of route-based control 3
Speed Prediction 4
Future Speed: Stochastic or Deterministic? Both! Impossible to know the exact future speed profile: driving is not deterministic! 2.5 2 1.5 1 But not completely random either Accel. [m/s 2 ].5 -.5-1 -1.5-2 DETERMINISTIC the driver follows an itinerary selected at the beginning of the trip OR the driver selects an itinerary in the navigation unit, and follows directions STOCHASTIC -2.5 2 4 6 8 1 Speed [km/h] Free flow: natural variations (accelerations vary, not always same speed) Interactions with other cars & environment 5
Maps / GIS Can Provide Information About a Given Itinerary ADAS-RP ADAS = Advanced Driver Assistance Systems RP = Research Platform Traffic pattern speed: average traffic speed for a given time/day Road slope: modeled with splines, not simply from GPS altitude data Speed limitations Position of traffic lights, stop signs, intersections, and other signs Category of road Number of lanes Etc. But not enough to predict fuel consumption! Vehicle Speed Distance 6
Chicago Real-World Data Provides Stochastic Aspect 4 3 2 1 25 2 15 1 25 5 5 2 1 15 2 25 15 1 2 3 4 5 1 5 1 2 3 4 5 27 travel survey for the Chicago Metropolitan Agency for Planning (CMAP) GPS loggers 267 households surveyed 1k vehicle trips 6M data points Acceleration [m/s 2 ] 3 2 1-1 -2 time density Top envelope Bottom envelope -3-4 Processing 2 4 6 8 1 12 14 Speed [km/h] 59% of points deemed valid (=1h) % of total 1.1.1.1.1 7
Speed (m/s) From Database to Actual Speed Profiles: Constrained Markov Chains TPM Initialization (t=, a=, v=) 1 1 5 5 2 4 1 5 1 2 3 4 1 2 3 4 Valid Real-World Micro-Trips. 5. 1. 2. 3. 5. 1. 2. 7 Transition Probability Matrix. 5. 1. 2.1 5 17. 16.5 16. 15.5 14.5 14. 13.5 P=.5 P=.15 P=.2 P=.3 P=.15 P=.1 P=.5 t-2 t-1 t t+1 Time (s) Markov Chains Random number generation Constrained Markov Chain Compute next state d>d target? v=v end? Metadata matches target? Speed Profile 8
Examples of Synthesized Speed Profiles Multiple stochastic speed profiles for the same target micro-trip One synthetic speed profile for one entire itinerary V V tgt max V avg act V avg t stop 9 Speed Limit 5 km/h 8 7 Speed (km/h) / Time (s) 6 5 4 3 Target Speed 32 km/h 2 1 1 2 3 4 5 6 Time (s) 9
Optimal Energy Management 1
Pontryagin s Minimum Principle System: one-mode power-split PHEV (similar to Toyota Prius PHEV) t influenced by control: Vehicle speed, torque demand decided by driver State of the system: battery state-of-charge (SOC) At any given time, for any given vehicle speed / wheel torque / battery power, one optimal operating point minimizing fuel consumption exists Battery power P b = command variable Constraints: final SOC is 3% PMP: Hamiltonian Optimal Command P b = argmin P b ( P f P b + r(t)θ P b P b ) Fuel Power Function of P b through optimal operation maps Equivalence Factor (EQF) Term close to 1 Battery Power Command In our study we make the assumption that EQF r t = r Optimal EQF: one that results in SOC=3% for the first time at the end of the trip 11
PMP Implementation in the Vehicle Controller Battery power Candidates Optimal Operating Points Fuel and battery power computation Minimization of cost function Filtering of the optimal power demand PMP w/ ICE Computation of corresponding torque/speed targets Optimal Speed/Torque Targets (HEV) Cost HEV ICE ON/OFF Logic Speed & Torque Targets ICE ON/OFF Cost EV EV Mode Speed/Torque Targets (EV) 12
Simulation Framework 13
Driver & Powertrain A forward-looking model of the Prius PHEV in Autonomie Driver presses on pedals Power-Split Hybrid-Electric (Toyota Prius Hybrid System) Vehicle energy management computes torque demands Powertrain = all components Torque (N.m) 15 1 5.35.35.35.3.35.3.3.25.25.25.2.2.2.15.15.15.5.1.5.1.5.1-5 2 3 4 Speed (rad/s) Components: dynamics + look-up tables from test data 14
Route Selection User can select route in HERE s ADAS-RP Route export plug-in ADAS-RP 15
Speed Prediction Input = Route from ADAS-RP Output = n speed profiles + grade 16
Route-Dependent Optimization Start SOC drops too fast Run EV+CS t SOC=3%? PMP Battery not used enough SOC tgt t end t SOC=3% <t end - δt? Increase EQF Run PMP Decrease EQF SOC tgt t end t SOC=3%? SOC end >SOC tgt +δsoc? Optimal SOC drop t SOC=3% <t end - δt? EQF Found SOC tgt t end 17
Actual Driving In the real-world, actual speed prediction will be different from prediction 1 Vehicle Speed (km/h) 1 Vehicle Speed (km/h) 5 5 1 1 5 p speed profiles for EQF optimization (before driving) 1 5 1 q speed profiles for actual driving 5 5 1 1 5 5 5 1 15 2 Distance (km) 5 1 15 2 Distance (km) 18
Large-Scale Evaluation 3 itineraries 8 Generations 3 SOC init 9 EQFs 1 5 1 5 Vehicle Speed (km/h) 1 EQF Optimization Start Run EV+CS 1 5 1 5 1 5 1 SOC tgt t end Increase EQF tsoc=3%? PMP tsoc=3%<tend - δt? Run PMP Decrease EQF SOCend>SOCtgt+δSOC? tsoc=3%? tsoc=3%<tend - δt? EQF Found 5 1 5 1 5 5 1 15 2 Distance (km) + 8 suboptimal values around optimal EQF 19
Example of Result (1 Itinerary, 1 generation) Cycle: Urban_2_3 Dist: 35.155 km Avg Spd: 43.7861 km/h 12 Speed (km/h) 1 grade (%)(x1) elev. (m) 8 6 4 2 SOC (%) 9 8 7 6 5 Ref Opt 2.53 2.54 2.55 2.56 2.58 2.59 2.6 2.61 Fuel(g) 4 35 3 25 2 15 Ref Opt 2.53 2.54 2.55 2.56 2.58 2.59-2 -4 4 1 2.6 2.61-6 3 5-8 5 1 15 2 25 3 Time (s) 2 5 1 15 2 25 3 Time (s) 5 1 15 2 25 3 Time (s) 15 15.5 Fuel Saving (%) 1 5-5 unadj. adj Fuel Energy (MJ) 15 14.5 14 13.5 Ref. Opt. Tgt SOC 2.6 2.59 2.58 2.57 2.56 2.55 EqF Fuel savings need to be SOC adjusted: final SOC in optimal case is always 3%, but it varies for reference case (stays in the [28,32] range) -1 13 2.54-15 2.52 2.54 2.56 2.58 2.6 2.62 2.64 EqF 12.5 8 8.5 9 9.5 1 Battery Energy (MJ) 2.53 2
Preliminary Results Show Strong Benefits (Best Case Scenario) 3 SOC =5% 25 SOC =7% SOC =9% 2 Adj. Fuel Savings (%) 15 1 5-5 -1 16 18 2 22 24 26 28 3 32 34 36 Trip Distance (km) 21
Conclusion Optimal energy management for xhevs theoretically requires full knowledge of duty cycle, which is not possible in the real-world A realistic and stochastic prediction can be achieved, through a combination of Markov chains and data from digital maps PMP is a convenient way of achieving optimal control in a real-world controller Efficacy depends on one tuning parameter, the equivalence factor (EQF) EQF depends on the future route A framework was designed to evaluate route-based control along with its uncertainties: Detailed powertrain model in Autonomie Vehicle speed prediction for a given itinerary Optimal route-based tuning Large-scale simulation to evaluate benefits in a broad range of situations Future work: Further statistical analysis: can the optimal EQF be inferred from simple parameters, or full simulation is needed? Make the controller adaptive, i.e. update EQF periodically (vs. simply at start of the trip) 22
Acknowledgement Funded by the Vehicle Technology Office Program Manager: David Anderson HERE (a kia company) provided free license for ADAS-RP Contact Dominik Karbowski (Principal Investigator): dkarbowski@anl.gov / 1-63-252-5362 Aymeric Rousseau (Systems Modeling and Control Manager): arousseau@anl.gov www.autonomie.net www.transportation.anl.gov The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory ("Argonne"). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract. DE- AC2-6CH11357. The U.S. Government retains for itself, and others acting on its behalf, a paidup nonexclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, by or on behalf of the Government.