Electric vehicle parameter identification

Abstract

This work describes a method developed to simplify and reduce the cost of the standardized testing required to measure the energy consumption of electric motorcycles before they are sold on the Swiss market. The robust algorithms for estimating four characteristic parameters of electric two-wheelers are based on a linear dynamics model that is considered to be as simple as possible yet as complex as necessary to characterize the vehicles. The model equations are investigated analytically for their ability to find unique solutions, and it is shown that multiple solutions may exist. The algorithms are tested for their ability to handle various data frequencies, levels of noise, and initial guesses. Ultimately it is found that these methods effectively enable the use of simulation models trained on real-world driving to run mandated standard test cycles in place of expensive dynamometer testing to estimate on-road energy use.