Modelling of EV penetration for an Indian state using Machine Learning Approach

Rising global warming level and its harmful impact on the environment have imposed a shift towards sustainable transportation systems. Electric vehicles (EVs) offer a promising solution to reduce greenhouse gas emissions and dependence on fossil fuels. This paper presents a model for predicting EV penetration in Uttarakhand, India, considering key parameters such as population, grid availability, altitude, and income. The model is developed using a Machine Learning algorithm know as Random Forest technique. To ensure realistic predictions, a dataset of 10,000 data points is generated using the probability density function, out of which the generated data is divided into training and testing data set having 70% and 30% ratio. The model is incorporated with random noise for enhanced robustness. The model assigns weights to the parameters, optimizing them to achieve a coefficient of determination close to 1. After training the model with 7,000 data sets and testing it with approximately 3,000 data sets, an R-square value of around 0.91566 is achieved, indicating a high level of accuracy. At last, using the developed model the total amount of energy required each day for charging is estimated helping the Distribution System Operators to calculate the amount of ancillary service required and the modifications need in the existing grid infrastructure.