Parameter optimization of electric vehicles according to driving behavior

The thesis aims to reduce the environmental and economic losses caused by the use of vehicles. To this end, the drivers are primarily divided into three classes: calm, normal and aggressive. Data were recorded from the test drives conducted with male and female drivers of different ages with vehicle tracking device and smartphone application. With this data, attribute extraction is made and classification accuracy of the drives with different attributes is examined. Support Vector Machine, K-Nearest Neighbor and a hybrid method using the Support Vector Machine and Markov Chain methods were used as classification algorithms, and the drives were divided into the correct classes with an accuracy of 98.9%, 93.3% and 92.2% respectively. The purpose of all these operations is to ensure the correct classification of the drivers from the available data and to optimize the electric vehicle for these drivers. Electric motor has been selected as the component to be optimized so that both battery and vehicle size can be changed. Motor power was determined for all drive classes as a result of optimization using Multiobjective Genetic Algorithm method. Lower engine power means lower battery, smaller car, less production costs, less carbon emissions. Greenhouse gas, which is harmful to nature, is released not only by the burned gasoline, but also during the production phase of the vehicle and the electricity used to charge the battery. With the regulation proposed by the study, economic and environmental important steps are taken by changing the preference of the car. Keywords: Driver classification, Parameter optimization, Electric vehicle, Support vector machine, Markov chain