Hybridization, Energy Management, and Vehicle Control

Introduction

A hybrid electric vehicle (HEV) combines at least two sources of traction power, commonly an internal-combustion engine and an electric drive system. Hybridization does not merely add a motor to a conventional vehicle. It creates a coordinated energy-conversion system in which the supervisory controller must decide, at every instant, how much power should be supplied by the engine, motor, battery, regenerative braking system, and auxiliary loads.

The central engineering problem is to satisfy driver demand safely and smoothly while minimizing fuel consumption, emissions, battery degradation, and component stress. This requires several control layers. Low-level controllers regulate torque, current, speed, and pressure. A supervisory energy-management system (EMS) determines operating modes and power split. Vehicle control coordinates propulsion, braking, transmission operation, thermal systems, and accessories.

This chapter develops rule-based, optimization-based, predictive, and learning-based control methods. It also explains practical implementation constraints, regenerative braking coordination, auxiliary-load management, and validation methods.

Learning Objectives

After completing this chapter, the reader should be able to:

1. Explain hybrid powertrain architectures and supervisory-control hierarchies.
2. Formulate vehicle power demand and power-balance relationships.
3. Compare rule-based, optimization-based, predictive, and learning-based EMS methods.
4. Coordinate engine, motor, transmission, battery, regenerative braking, and auxiliary loads.
5. Apply state-of-charge constraints, efficiency maps, and operating limits in control design.
6. Evaluate energy-management strategies using fuel, emissions, drivability, battery, and safety criteria.

1. Hybrid Powertrain Architecture and Control Hierarchy

Common architectures are:

• Series hybrid: the engine drives a generator; the electric motor alone drives the wheels.
• Parallel hybrid: engine and motor can both deliver torque mechanically to the wheels.
• Series-parallel hybrid: power can flow through either electrical or mechanical paths.
• Plug-in hybrid electric vehicle (PHEV): a larger battery permits substantial grid-charged electric operation.
• Mild hybrid: a relatively small motor assists the engine, often at a belt or crankshaft interface.

A practical control hierarchy contains three levels:

The supervisory controller receives vehicle speed, accelerator and brake requests, battery state of charge (SOC), temperatures, road information, and component limits. It sends commands to the engine controller, inverter, transmission, brake controller, and thermal-management system.

A useful design principle is separation of objectives. The supervisory layer optimizes energy and mode selection, while fast inner loops ensure that requested torque and current are accurately achieved. A supervisory controller should not directly manipulate switching signals or actuator currents unless the entire control architecture is designed for that purpose.

2. Vehicle Power Demand and Energy Balance

The tractive force required at the tire-road interface can be approximated by

where F_t is tractive force (N), F_a is aerodynamic drag, F_r is rolling resistance, F_g is grade force, m is vehicle mass (kg), and a is longitudinal acceleration (m/s²).

For a vehicle traveling at speed v,

where ρ is air density (kg/m³), C_d is drag coefficient, and A_f is frontal area (m²). Rolling resistance is commonly approximated by

where g is gravitational acceleration, C_rr is the rolling-resistance coefficient, and θ is road angle. The grade force is

These expressions assume a rigid vehicle, known road angle, and longitudinal motion. They neglect detailed suspension, tire-slip, and transient aerodynamic effects.

Wheel power is

and the approximate battery or fuel-converter power depends on the selected operating mode and conversion efficiencies. During propulsion, if motor efficiency is η_m and inverter efficiency is η_i, battery discharge power required for motor output P_m is approximately

where P_aux is auxiliary electrical power. During regenerative braking, the battery receives approximately

subject to battery charge-power limits. The symbols should not be interpreted as universal constants: efficiencies vary with speed, torque, temperature, and direction of power flow.

Battery SOC is often represented by a coulomb-counting model:

for discharge current I_k defined as positive, nominal capacity Q_n in ampere-hours, sampling interval ∆t in hours, and discharge efficiency η_d. A more detailed model uses separate charge and discharge efficiencies and an open-circuit-voltage/resistance equivalent circuit. Coulomb counting accumulates sensor bias; therefore, practical systems periodically correct SOC using voltage, model observers, or calibrated operating conditions.

3. Energy-Management Objectives and Constraints

The EMS seeks a feasible compromise among competing objectives:

• low fuel consumption and greenhouse-gas emissions;
• low regulated emissions;
• acceptable battery SOC and battery aging;
• smooth torque delivery and shift quality;
• low noise and vibration;
• component-temperature control;
• safety and fault tolerance.

A general finite-horizon cost function can be written as

Here, ṁ_f is fuel-mass flow rate, E_em represents emissions or an emissions penalty, C_deg is a battery-degradation cost, e_drv is a drivability error, and the w terms are design weights. The terminal SOC penalty discourages a controller from ending a test with an artificially depleted battery.

Typical constraints include

and actuator limits such as engine torque, motor torque, inverter current, brake pressure, and transmission speed ratio. Constraints may be hard, meaning violation is unacceptable, or soft, meaning violation is penalized but permitted temporarily under carefully controlled conditions.

4. Rule-Based Energy Management

Rule-based control is widely used because it is transparent, computationally inexpensive, and straightforward to calibrate. Rules may be deterministic, fuzzy, or state-machine based.

A typical state machine includes:

1. Electric launch: use motor torque when battery SOC and temperature permit.
2. Engine start: start the engine when demanded power, speed, or SOC conditions exceed thresholds.
3. Engine assistance: combine engine and motor torque during high demand.
4. Engine-only cruising: operate the engine near an efficient region when electrical assistance is unnecessary.
5. Charging mode: load the engine to charge the battery when SOC is low and charging is permitted.
6. Regenerative braking: allocate braking demand first to regenerative braking, then friction brakes as required.
7. Engine stop: stop fuel injection during suitable low-load or stationary conditions.

A simple power-split rule might command motor assistance when

and command engine charging when SOC falls below a lower threshold. In practice, hysteresis is necessary. For example, engine start and stop thresholds should differ so that small fluctuations do not cause repeated starts and stops.

Fuzzy control replaces sharp thresholds with linguistic rules such as “IF SOC is low AND power demand is medium, THEN engine power is high.” Membership functions convert measured variables into degrees of membership. Fuzzy rules can represent expert calibration knowledge but may be difficult to verify exhaustively.

Advantages: low computation, predictable behavior, easy certification and calibration.

Limitations: performance depends strongly on thresholds, rules may not generalize to all drive cycles, and interactions among many variables can become difficult to manage.

5. Optimization-Based Energy Management

Optimization-based methods explicitly evaluate operating cost. The most fundamental benchmark is dynamic programming (DP). Given a known drive cycle, DP searches over possible control sequences and SOC states to minimize total cost. It can provide a near-global optimum for a discretized model, but it requires future knowledge and can be computationally expensive. Thus, DP is mainly used for design benchmarking and calibration rather than direct real-time control.

Equivalent consumption minimization strategy (ECMS) converts electrical energy use into an equivalent fuel cost. A simplified instantaneous objective is

where s is an equivalence factor with units of fuel mass per joule, and P_b,eq represents battery power with sign and efficiency conventions carefully defined. Discharge power generally increases equivalent fuel consumption, while charging power may reduce it. The equivalence factor is calibrated or adapted so that SOC remains near its desired trajectory.

Pontryagin’s minimum principle (PMP) uses a Hamiltonian containing instantaneous cost and a costate multiplied by the state dynamics. In a simplified form,

where x is the state, u is the control, λ is the costate, L is instantaneous cost, and f describes state evolution. The control minimizing the Hamiltonian is selected subject to constraints. PMP can produce real-time strategies, but costate adaptation and model accuracy are important.

Model predictive control (MPC) repeatedly predicts future behavior over a finite horizon, solves a constrained optimization problem, applies only the first control action, and then repeats the process with new measurements. MPC can coordinate engine operating points, transmission ratio, battery SOC, and braking while explicitly enforcing limits. Its disadvantages are computational cost, sensitivity to prediction quality, and the need for a suitable reduced-order model.

6. Predictive and Learning-Based Control

Connected vehicles can obtain preview information from navigation maps, traffic signals, elevation data, weather services, and traffic prediction. Preview allows the controller to reserve battery energy for an upcoming hill, congestion region, or zero-emission zone. Prediction must be treated probabilistically because traffic and driver behavior are uncertain.

Learning-based methods include supervised learning, reinforcement learning (RL), and hybrid model-learning methods. In RL, an agent selects an action based on a state and receives a reward. A conceptual reward may be

RL can discover control policies for complex nonlinear systems, but an unconstrained learned policy cannot be assumed safe. Deployment should use action limits, safety filters, fallback rules, offline validation, and monitoring. A practical architecture is a learning-based optimizer supervised by a rule-based safety layer.

Learning methods face challenges involving distribution shift, explainability, training-data coverage, cybersecurity, and certification. They are most useful when combined with physical models rather than used as unbounded replacements for them.

7. Coordinated Control of Propulsion and Transmission

The engine typically has a high-efficiency region in its speed-torque map. The EMS should therefore avoid operating the engine at very low load when the battery or motor can provide the required power, and should avoid excessive engine loading that causes poor emissions or drivability. In a parallel hybrid, the motor can shift the combined operating point toward a more efficient region.

Transmission control must coordinate gear selection with engine and motor torque. A lower gear may improve acceleration but increase engine speed and losses. A higher gear may reduce cruising losses but limit available wheel torque. During a shift, torque intervention is required to reduce jerk and clutch energy. The supervisory controller should coordinate:

• engine torque reduction or increase;
• motor torque fill;
• clutch pressure;
• transmission input and output speeds;
• shift timing;
• battery power limits.

Wheel torque can be represented by

where T_e and T_m are engine and motor torques, i_g is the selected gear ratio, i₀ is the final-drive ratio, and η_d is driveline efficiency. This expression assumes a compatible mechanical coupling and neglects transient inertia torque; more complete models include rotational inertias and shaft dynamics.

8. Regenerative Braking and Brake Blending

During deceleration, the motor operates as a generator and converts part of the vehicle’s kinetic and potential energy into electrical energy. Regeneration is limited by motor speed, inverter current, battery SOC, battery temperature, tire-road adhesion, and requested deceleration.

The controller first determines total braking demand, then allocates it between regenerative and friction braking. If total required wheel braking torque is T_req, a feasible allocation is

with

where all torques are expressed using a consistent wheel-side sign convention. Friction brakes supply the remainder. Brake blending must preserve the driver’s requested deceleration and maintain stability under changing adhesion.

Important practical requirements are:

• smooth transition between regenerative and friction braking;
• anti-lock braking compatibility;
• stable operation on split-μ surfaces;
• predictable pedal feel;
• full friction-brake availability when the battery is full or cold;
• energy recovery without compromising safety.

Regeneration is not always optimal. At very low speed, conversion losses may exceed recoverable energy. In addition, deliberately increasing engine braking or charging the battery can be inferior to coasting, depending on efficiency and future demand.

9. Auxiliary Loads and Thermal Management

Heating, ventilation, air conditioning, pumps, fans, lighting, steering, braking, and information systems consume energy. In an HEV, auxiliary loads can significantly affect fuel use, particularly at low vehicle power. Cabin heating may require engine waste heat, a positive-temperature-coefficient heater, or a heat pump. Cooling systems influence battery power capability and life.

The EMS should therefore include

where each term is electrical power measured at a consistent bus boundary. Load coordination may include preconditioning while grid-connected, reduced compressor operation during acceleration, and thermal preparation before high-power charging or climbing.

Thermal constraints should not be treated as afterthoughts. A controller that minimizes immediate fuel consumption but repeatedly overheats the battery or inverter is not an acceptable controller. Battery aging models may include temperature, SOC, and current stress; their parameters require experimental validation.

10. Worked Example: Power Demand and Regenerative Braking

Consider a 1,600 kg vehicle traveling at 20 m/s on a level road. Let C_d = 0.29, A_f = 2.2 m², ρ = 1.2 kg/m³, C_rr = 0.012, and g = 9.81 m/s². Assume drivetrain efficiency of 0.90 and auxiliary demand of 1.0 kW.

Aerodynamic drag is

Rolling resistance is

At constant speed, acceleration and grade force are zero. Therefore,

The approximate electrical power required is

This calculation is a steady-state estimate. It excludes tire losses not represented by C_rr, transient acceleration, accessory variation, and detailed efficiency-map effects.

Now suppose the vehicle decelerates from 20 m/s to rest. Its initial kinetic energy is

If 65% of this energy is recoverable after motor, inverter, and battery limitations, recovered energy is

The value is modest compared with a traction battery’s total energy, illustrating why repeated urban braking events—not one isolated event—determine the importance of regeneration.

11. Illustrative Case Study: Supervisory Control for a Plug-In Hybrid Bus

Background

A city operates a plug-in hybrid bus on a route containing frequent stops, a steep outbound grade, and a low-emission zone near the city center. The bus has an engine-generator set, a traction motor, a battery, regenerative braking, and electrically driven auxiliaries.

Problem

A simple SOC-threshold controller preserves battery energy poorly. It uses electric power on the outbound grade, reaches the low-emission zone with insufficient SOC, and frequently switches the engine on and off near threshold values. Passengers experience torque disturbances, while battery temperature rises during repeated high-power operation.

Analysis

The route has predictable elevation and stop locations. The controller therefore has useful preview information. The principal requirements are zero-emission operation in the designated zone, sufficient power for the grade, smooth engine transitions, and safe battery temperature.

Approach

A constrained MPC strategy is implemented with a 60-second prediction horizon. Its state variables include SOC, battery temperature, vehicle speed, and engine operating mode. The cost function penalizes fuel use, engine starts, SOC deviation, battery temperature, and deviation from the requested acceleration. Constraints impose battery current limits, minimum SOC at the zone entrance, engine power limits, and maximum regenerative braking torque.

A rule-based fallback controller is retained for sensor faults, failed predictions, or optimization timeout. Regenerative braking is prioritized at each stop, but friction braking is blended whenever battery charge power or tire adhesion is limiting. Auxiliary power is included in the prediction model, and cabin preconditioning occurs while the bus is connected to the grid.

Outcome

In an illustrative simulation—not a measured field trial—the predictive controller reaches the low-emission zone with the required SOC reserve, reduces unnecessary engine cycling, and maintains smoother battery temperature. Its benefit depends on accurate route preview and calibration. If traffic conditions differ substantially from the prediction, the fallback rules and terminal SOC constraint prevent unsafe or impractical behavior.

Learning Points

1. Preview information is valuable only when integrated with constraints.
2. SOC targets should depend on future power demand, not merely present SOC.
3. Safety and fallback control must remain independent of optimization success.
4. Auxiliary and thermal loads materially affect supervisory decisions.
5. Simulation results must not be presented as field validation without measurement evidence.

12. Validation, Calibration, and Professional Practice

A control strategy should be tested progressively:

1. Model-in-the-loop: evaluate algorithms using mathematical plant models.
2. Software-in-the-loop: test production-oriented software behavior.
3. Hardware-in-the-loop: connect real controllers to real-time plant simulation.
4. Component testing: validate engines, motors, batteries, brakes, and transmissions.
5. Vehicle testing: assess performance over standardized and representative routes.

Important metrics include fuel consumption, electrical energy use, SOC deviation, regenerative-energy fraction, emissions, battery temperature, engine starts, shift quality, longitudinal jerk, brake-pedal consistency, and computation time. Testing should include cold and hot conditions, high and low SOC, payload variation, steep grades, low-adhesion surfaces, sensor faults, communication loss, and actuator saturation.

Calibration is a multi-objective activity. Improving fuel consumption may increase engine noise or battery degradation. A robust calibration documents assumptions, test conditions, software version, parameter sources, and acceptance criteria.

Cybersecurity and functional safety are essential. Commands from connected services should be authenticated and bounded. A learned or predictive controller must fail safely, and safety-critical braking functions require independent monitoring and validated fallback behavior. Data collection should respect privacy when driver, location, or behavioral data are recorded.

13. Contemporary Developments and Design Best Practices

Current development trends include centralized vehicle computers, high-fidelity digital twins, cloud-supported fleet optimization, silicon-carbide inverters, battery-health-aware control, and predictive thermal management. These technologies can improve efficiency but also increase software complexity and verification burden.

Recommended practices are:

• define clear supervisory interfaces and sign conventions;
• use physically meaningful states and validated component maps;
• enforce actuator, thermal, SOC, and safety constraints explicitly;
• include terminal SOC or charge-sustaining requirements in comparisons;
• use hysteresis or state estimation to prevent mode chattering;
• separate optimization from safety enforcement;
• validate across broad environmental and traffic conditions;
• report both energy benefits and degradation or emissions consequences;
• retain interpretable fallback strategies;
• document calibration assumptions and uncertainty.

Chapter Summary

Hybrid vehicle control is a coordinated, hierarchical activity. The supervisory EMS determines operating modes and power split, while lower-level controllers regulate torque, current, braking, and thermal actuators. Vehicle power demand is determined by aerodynamic drag, rolling resistance, grade, and acceleration. Battery SOC, power limits, temperature, and aging impose important constraints.

Rule-based strategies are transparent and practical; optimization-based methods can improve efficiency but require models and computation; predictive control exploits route and traffic information; and learning-based methods can address nonlinear complexity only when combined with safety limits, validation, and fallback control. Effective systems coordinate the engine, motor, transmission, regenerative braking, friction brakes, and auxiliary loads. Evaluation must include fuel, emissions, drivability, battery health, thermal behavior, safety, and robustness—not fuel consumption alone.

Key Terms

Hybrid electric vehicle: Vehicle using two or more traction-energy sources.

Supervisory controller: Controller that selects operating modes and power allocation.

Energy-management system: System that manages power flow to meet performance and efficiency objectives.

State of charge (SOC): Estimated available battery charge relative to a reference capacity.

Series hybrid: Architecture in which the engine generates electricity and the motor drives the wheels.

Parallel hybrid: Architecture in which engine and motor can mechanically propel the wheels.

Power split: Allocation of demanded traction power among energy converters.

Regenerative braking: Recovery of braking energy through generator operation of the motor.

Brake blending: Coordinated allocation of braking demand between regenerative and friction brakes.

ECMS: Equivalent consumption minimization strategy.

Dynamic programming: Optimization method that searches decisions over states and time.

Model predictive control: Receding-horizon control using predictions and constraints.

Costate: Variable representing the marginal value of a dynamic state in optimal control.

Hysteresis: Separation between switching thresholds used to prevent repeated mode changes.

Auxiliary load: Non-traction electrical or mechanical power demand.

Driveability: Perceived quality of vehicle response, smoothness, and predictability.

Charge sustaining: Operation that maintains battery SOC within a prescribed range.

Fallback controller: Safe alternative controller used when the primary controller is unavailable.

Battery degradation: Loss of battery capacity or power capability due to use and aging.

Terminal SOC constraint: Requirement or penalty associated with SOC at the end of an optimization horizon.

Review Questions

Short Answer Questions

1. What is the primary role of a supervisory energy-management controller?
2. Why is hysteresis used in engine start-stop logic?
3. Identify two constraints that limit regenerative braking.
4. What is the purpose of a terminal SOC penalty?
5. Distinguish a series hybrid from a parallel hybrid.

Long Answer Questions

1. Explain the control hierarchy of a hybrid electric vehicle and describe the information exchanged between its layers.
2. Compare rule-based control, ECMS, dynamic programming, and MPC in terms of computation, optimality, transparency, and practical deployment.
3. Derive the principal components of longitudinal tractive force and discuss the assumptions behind each component.
4. Explain how engine, motor, and transmission control can be coordinated during a gear shift.
5. Discuss the safety, cybersecurity, and validation requirements for learning-based energy management.

Analytical / Application Questions

1. A vehicle has low SOC, high battery temperature, and a steep upcoming grade. Propose a supervisory strategy and justify the timing of engine operation, battery use, and thermal management.
2. Design a brake-blending logic for a vehicle whose battery charge-power limit changes rapidly with temperature and SOC. Identify the safety fallback behavior.
3. Compare two controllers over a drive cycle when one ends with substantially lower SOC than the other. Explain how to make a fair fuel-consumption comparison.

Multiple Choice Questions with Answers

Question 1

Which controller layer normally selects the engine–motor power split?

A. Supervisory controller

B. Inverter switching controller

C. Wheel-speed sensor

D. Brake-pad temperature sensor

Correct Answer: A. Supervisory controller

Question 2

For level-road constant-speed travel, which term is zero in the simplified tractive-force model?

A. Aerodynamic drag

B. Rolling resistance

C. Grade force

D. Vehicle mass

Correct Answer: C. Grade force

Question 3

What is the main purpose of hysteresis in hybrid mode logic?

A. Increase battery voltage

B. Prevent rapid repeated switching

C. Eliminate aerodynamic drag

D. Measure fuel composition

Correct Answer: B. Prevent rapid repeated switching

Question 4

Dynamic programming is most commonly used in hybrid control to:

A. Provide a benchmark using known future drive-cycle information

B. Measure tire pressure directly

C. Replace the mechanical brake system

D. Increase motor winding resistance

Correct Answer: A. Provide a benchmark using known future drive-cycle information

Question 5

In regenerative braking, friction brakes are required when:

A. The battery and motor can accept unlimited power

B. Requested braking exceeds regenerative capability

C. Vehicle speed is exactly constant

D. Auxiliary loads are zero

Correct Answer: B. Requested braking exceeds regenerative capability

Question 6

A vehicle travels at 20 m/s with 1,600 kg mass. Its kinetic energy is:

A. 32 kJ

B. 160 kJ

C. 320 kJ

D. 640 kJ

Correct Answer: C. 320 kJ

Question 7

Which method repeatedly optimizes over a future horizon and applies only the first action?

A. Model predictive control

B. Coulomb counting

C. Static gear selection

D. Open-loop ignition timing

Correct Answer: A. Model predictive control

Question 8

Why should auxiliary loads be included in energy management?

A. They can materially affect total electrical power demand

B. They determine tire-road friction exclusively

C. They eliminate battery temperature limits

D. They are always larger than traction power

Correct Answer: A. They can materially affect total electrical power demand

Question 9

Which condition generally reduces the allowable regenerative braking power?

A. Battery charge-power limitation

B. Low cabin lighting demand

C. High windshield transparency

D. Reduced audio volume

Correct Answer: A. Battery charge-power limitation

Question 10

What is a principal risk of deploying an unconstrained learning-based controller?

A. It cannot represent any nonlinear behavior

B. It may select unsafe actions outside its training distribution

C. It always requires a mechanical transmission

D. It necessarily disables regenerative braking

Correct Answer: B. It may select unsafe actions outside its training distribution