Optimal EV Charging Infrastructure

Optimisation

The notebook was built using GitHub

Introduction The mobility sector accounts for around 18% of the carbon dioxide emissions from burning fuel. A considerable market share of electric vehicles (EVs) running on clean power is key to mobility decarbonisation. As the number of EVs on our roads continues to increase, the demand for charging is following suit. Today around 10% of drivers are choosing EVs when buying a new car, and this is expected to grow rapidly.

We need accelerated technology and infrastructure development to support EV market growth. While EV drivers are looking for a charging experience that is as fast and comfortable as possible, the main obstacle to mobility decarbonisation is inconvenience caused by charging infrastructure and charging time Proper placement of charging points can alleviate this problem. It is a subject of current research and business interest simultaneously and has immense implications for the adoption of EVs, including:

lowering range anxiety among EV owners,
optimal utilisation of EV charging points,
minimal travel time and waiting time for EV owners. In addition, the challenges around this problem are evolving as EVs penetrate different geographies. For example, the EV charging placement in a U.S. city has different nuances than in a suburban town in India.

The demand for solving this problem at different geographies and scales will increase exponentially in the next decades as EVs spread from cities and urban areas to villages.

Problem Statement:

To optimally place EV charging stations so that the configuration remains robust to demographic changes.

Notations and Constraints

1. Modeling Approach

The project output demands the following steps:

Forecast the EV demand for each demand point
Estimate the distance to the nearest supply point
Create the demand-supply matrix
Given the constraints, formulate an LP problem to minimise the overall cost

1.1. Forecast the EV demand for each demand point

Exploratory dat analysis suggests, the current year’s demand is dependent on previous year.
Also, the average change in EV demand over the years is highy positively correlated with current years demand
Using the above two features, demand of EV has been forecasted using Linear Regression with an MAE of 0.34

1.2. Estimate the distance to the nearest supply point

Distance estimation from the demand point to the supply point is a function of magnitude.
Applied Euclidean Distance to estimate distance between each demand and supply index
This resulted in a demand-supply matrix of mxn dimensions. i.e. 4096*100
Simply put, distance from every demand point to every supply point was calculated

1.3. Formulate an LP problem to minimise the overall cost

- Defining the objective function - adding all costs

- Defining the decision variables

(a) the demand fulfilled at every demand index by every supply index, (b)optimal number of scs and (c)optimal number of fcs

- Defining the constraints

(a) Sum of SCS and FCS must be less than or equal to the total parking slots available at each supply point
(b) SCS and FCS must increase or stay constant year-on-year at each supply point (c) Forecasted demand at each demand point must exactly be satisfied

- Solving the LP problem