A Comprehensive Study of Control Methodology for Plug-in Hybrid Electric Vehicles.

摘要

Vehicle electrification has been well acknowledged as the most promising technology to achieve the efficient and clean transportation. Plug-in hybrid electric vehicles (PHEVs) are one of the major targets due to its potential of fuel displacement and extended driving range. However, the different operation cost of the two energy sources, electric energy from the utility grid and fuel energy, makes the drivetrain control problem more complex. This dissertation is dedicated to study the design and implementation of control methodology for PHEVs.;The drivetrain design is delivered first. Based on the targeted vehicle dynamic performances, each component of the hybrid drivetrain are properly sized and later applied in the simulation. Then, the driving pattern identification using LVQ neural network is proposed. It uses the statistical feature of the short term historical speed profile as the input. Instead of trying to identify and categorize driving cycles from detailed characteristics, only the driving patterns (highway or urban stop-and-go) are targeted.;Since the cost of utilizing the two energy sources (engine and motor) significantly differs on the highway and urban driving, the pattern identified can serve as guidance for the "smart" utilization of the "cheap and clean" electric energy. By extracting the results from global optimization, a deterministic energy management strategy is established subsequently. With the given trip distance, it switches the energy source between the battery and engine based on the recognized driving pattern such that the both the all-electric capability and the high average system efficiency are achievable. Additionally, in order to exploit the full potential of fuel displacement, lowest SOC is expected by the end of the trip. To this end, the remaining distance-to-go and AER are monitored so that the operation can shift to the allelectric when necessary to deplete the battery by the end of the trip.;One of the most challenging problems in the control of hybrid drivetrain is to determine the optimal power split between the engine, traction motor, and generator if applicable. As a result, in the next step, an optimization based power management strategy is developed. Using the innovatively proposed algorithm the normalized comprehensive energy loss of each power split is available to be computed. It is then employed as the criteria for determining the optimality of each operation. The power split with the minimum energy loss is believed being able to achieve better fuel economy. Moreover, the proposed normalized comprehensive energy loss can also serve as an indicator of customers' behavior, similar to the MPG in conventional vehicles.;Real time implementation depends on how fast the global minima can be located. For the drivetrain like series-parallel, the solution domain for a power demand is an unknown surface. Searching for the optimal solution could be time-consuming. Based on the characteristics of the solution domains, algorithms such as Particle Swarm Optimization (PSO) and Dividing Rectangles (DIRECT), as well as the look up table method are investigated. By trading off between the accuracy and speed, outcomes show that look up table method is the most appropriate approach thus is eventually applied.;Finally, the result from simulation is evaluated by comparing with the global optimal strategy. The comparison shows that the proposed control strategy can achieve comparable overall fuel economy with the global optimization, but is able to be implemented in real time control.