Modeling and Control of Hybrid Electric Vehicles Tutorial Session Ardalan Vahidi And Students: Ali Borhan, Chen Zhang, Dean Rotenberg Mechanical Engineering, Clemson University Clemson, South Carolina 2008 American Control Conference Seattle June 11, 2008 1/30
The Tutorial Session Overview on Modeling and Control of Hybrid Electric Powertrains (Ardalan Vahidi, Clemson University) Understanding Opportunities for Energy Management Control in HEVs through Degree of Freedom Analysis (Tony Phillips and Ming Kuang, Ford Motor Company) Configuration, Sizing and Control of Power Split Hybrid Vehicles (Huei Peng, University of Michigan) The Role of System Theory in Reducing Energy Losses in Hybrids (Lino Guzzella, ETH, Zurich) Design and Control of a Renewable Energy Based Eco System With Plug In/V2GHybrid Electric Vehicles (Georgio Rizzoni, Ohio State University) 2/30
Overview Why Hybrids Improve Fuel Economy Configurations of a Hybrid Powertrain How Systems and Controls Help Supervisory Control Strategies Some Recent Trends 3/30
Different Types of Hybrids Main Propulsion: Gasoline or Diesel Engines, Fuel Cells Assist Propulsion: Motors, Generators, Hydraulic Pumps Energy Storage: Batteries, Ultracapacitors, Flywheels, Hydraulic Accumulators The main focus of this talk is on gasoline engine electric hybrid electric vehicles with battery/ ultracapacitor storage with the purpose of improving fuel economy. Main Configurations: Series, Parallel, Powersplit 4/30
Well to Wheel Efficiencies Available Online: http://www.toyota.co.jp/en/tech/environment/hsd/pdf/a_guide_to_hsd.pdf 5/30
How Does Hybridization Help? Reducing braking losses through regeneration Possibility of clutch less operation thus reduced losses Potential for downsizing the engine Enabling operation of the propulsion system at its more efficient operating regions Allowing engine shut down during idle 6/30
Role of Control Engineering Engine downsizing, removing the clutch, and hybrid system configurations are mainly design considerations (system engineering perspective is still necessary). Regenerative braking is free energy, so no sophisticated high level control needed. Low level control still needed to optimize regeneration of electrical machine. Control engineering is critical in determining when to run the engine and at what power split ratio. The problem is very complex due to its dynamic nature and uncertainty about future driving conditions. 7/30
Engine Fuel Consumption and Efficiency Higher Fuel Rates Engine Max. Efficiency Curve Lower Fuel Rates 8/30
Series Configuration of HEVs The engine and the driveline are decoupled. Drawbacks: electrical path losses; motor/generator should be sized for max power. 9/30
Parallel Configuration of HEVs Engine power (or torque) can be decoupled from demand power. The engine speed is determined by the vehicle speed (and the transmission). Less power conversion loss compared to series and therefore less power losses. 10/30
Power Split or Series Parallel Configuration Combines advantages of parallel and series configuration, the output power can be delivered from both electrical and mechanical paths. Degrees of freedom are battery power and engine speed or torque. 11/30
The Challenge of Energy Management In an ideal world where the electrical path had an efficiency of 1, then the battery could be thought of a free buffer and charged/discharged at any time. The fuel economy optimization problem would be simplified to running the engine at most efficient line + charge sustaining considerations. In reality each charge/discharge results in losses: Battery should be used conservatively and predictively. Running the engine at most efficient line is NOT equivalent to running the hybrid most efficiently! Another concern is battery s life which is sensitive to cycling and depth of discharge. (less of a problem for an ultracapacitor). 12/30
The Control Objective Minimize fuel use and/or reduce emission without compromising the drivability of the vehicle. Additional Limits: Charge sustaining is needed for vehicle certification. Cycling of the energy storage should be limited to prolong battery life. T demand T engine Velocity SOC Gear Power Management Strategy T motor T generator T brake 13/30
Backward Looking or Quasi Static Models (Non Causal) Parallel Hybrid Example: Vehicle Driveline Engine v cycle T traction T engine Fuel rate ω wheel ω eng F road ω motor T motor Voltage Current SOC Electric Machine Battery + Inverter ADVISOR uses a backward simulation model. 14/30
Forward Looking or Dynamic Models (Causal) Parallel Hybrid Example: Vehicle Driveline Engine v T traction T engine Fuel rate ω wheel ω eng F road ω motor T motor Voltage Current SOC Electric Machine Battery + Inverter PSAT uses a forward simulation model. 15/30
Parallel Hybrid Example: Models of Hybrid Powertrain Components Vehicle Driveline Engine v In forward looking approach velocity is modeled as a dynamic state. T traction ω wheel torque/speed couplers: Static modeling okay. torque converter: Speed dynamics important in transients. T engine ω eng Fuel rate statically mapped from torque and speed. Torque dynamics fast and can be ignored. F road ω motor T motor Efficiency statically mapped from torque and speed. Torque dynamics fast and can be ignored. Voltage Current Power electronics: No need to model dynamics. Battery: State of charge should be modeled as a dynamic state. SOC Electric Machine Battery + Inverter 16/30
Energy Management Control strategies used: Rule based Fuzzy Logic/Neural Network Dynamic Programming Analytic Optimal Control Variable Structure Control (Sliding optimal control) ECMS, Adaptive ECMS Model Predictive Control 17/30
Rule Based Methods: An Example Assist Mode: Run Motor + Engine High optimal curve Medium optimal curve Run the engine at next higher optimal curve, charge the battery Low optimal curve Pure Electric Region Run in Pure Electric Mode The pure electric threshold is made a function of SOC for charge sustaining. 18/30
Pros and Cons of Rule Based Intuitive Simple to implement Fuel economy sensitive to threshold curves Charge sustaining is cycle dependent Fundamentally the optimal curve will depend on FUTURE charging/discharging conditions 19/30
Optimization Based Method The goal is to determine the power split that minimizes fuel use over an entire cycle, i.e. find P batt that minimizes: Ensuring that the charge is sustained: subject to the system dynamics. When SOC is the only state: and several inequality constraints, for example motor and engine torque limits. J t = t 0 f m f dt SOCt ( ) = SOCt ( ) f d V V 4R P SOC = dt 2C R 2 oc oc batt batt batt 0 batt 20/30
Dynamic Programming Assumes known drive cycle and solves the fuel minimization problem backwards in time based on Bellman s optimality principle. State of Charge 0.8 0.7 0.6 time t 0 t 1 t k t k+1 t f minimum fuel path 21/30
DP is Non Causal Unfortunately future driving conditions are normally highly uncertain: Driver behavior Road profile Traffic conditions Still DP can be used to: Evaluate an upper bound to fuel economy potential of a hybrid Learn from and extract rules for real time control When future driving cycle can be estimated using telematics or for vehicles that have repeatable cycles. Stochastic DP has also been proposed with a probablistic driver model. 22/30
Optimal Control: Let s Simplify J t f = t 0 m f dt d V V 4R P SOC = dt 2C R 2 oc oc batt batt batt batt Find u=p batt that minimizes J. Without inequality constraints, the Hamiltonian is: and V V 4R P Hut (, ) = m f + λ( t) 2CbattR batt 2 oc oc batt batt d H( u, t) λ = = 0 dt SOC In the case of a battery, this renders a constant λ. The constant optimal λ is the one that ensures charge sustaining. AGAIN DEPENDENCE ON FUTURE CYCLE! A constant estimate λ 0 can be obtained by inferring the driving conditions; but fuel economy and final charge very sensitive to this choice. 23/30
Equivalent Consumption Minimization Strategies: ECMS Instead of the charge sustaining constraint, use of electrical energy is directly penalized in an instantaneous cost function using a fuel equivalence function Ф: ( ) J = m H +ϕ P f LHV batt The challenge is approximating the equivalence function which depends on energy conversion efficiencies in the future. One popular approach is evaluation of Ф based on the average energy paths from fuel to storage of electrical energy and viceversa. J = m H + s P f LHV batt 24/30
Minimize Model Predictive Power Management i+ P ( ) 2 min J = Qm + Q SOC SOC u k= i 2 1 f 2 0 Subject to system dynamics and several constraints. Use a receding horizon approach. Can be translated to a: Nonlinear program, dynamic program or simplified to a quadratic or linear program. MPC suitable for handling complexity of next generation hybrids: Multiple degrees of freedom requires more systematic optimization Pushing system components to their limits: constraints More far sighted than ECMS still causal and real time implementable Ease of integration of future road data 25/30
MPC: Preliminary Results Dean shows reduction in fuel use with MPC over his rule-based algorithm for an ultracapacitor parallel hybrid. (Results presented earlier.) Ali shows a more flexible and systematic control design with MPC for multi- DOF hybrids. (Results to appear next year.) Sponsored by Ford Motor Company 26/30
Still a Next Step: Use of Future Road Information Preliminary Results: Chen shows up to 4% additional reduction in fuel use when future road grade known in highway driving, depending on the terrain. A modified rulebased was used. (Details in the next ACC) Sponsored by Intermap Technologies 27/30
Ultracapacitor Assisted Powertrains BMOD0140 Maxwell ultracapacitor module: Capacitance: 140F Voltage: 48 Volts Mass~13kg Energy: Only 160 kj Power: Up to 30~60kW instantaneously Demonstrated up to 15% reduction in fuel use for city cycle with 2 ultracapacitor modules shown above and a 40kW induction motor. D. Rotenberg, A. Vahidi, and I. Kolmanovsky, Ultracapacitor Assisted Powertrains: Sizing, Modeling and Control, and The Impact on Fuel Economy 2008 ACC. 28/30
Finally My Favorite Hybrid: The Electric Bicycle Question: How much of the potential energy can be recovered by regeneration? mgh Undergrads: Maria, Carl, and Seneca riding the E Bike. 29/30
Main References L. Guzzella, and A. Sciarretta, Vehicle Propulsion Systems, Springer 2005. M. Ehsani, Y. Gao, S. Gay, and A. Emadi, Modern Electric, Hybrid Electric, and Fuel Cell Vehicles, CRC Press 2005. J. Larminie, and J. Lowry, Electric Vehicle Technology Explained, Wiley 2003. And several technical papers. 30/30