Electric Vehicle Charging Stations Placement Optimization in Vietnam Using Mixed-Integer Nonlinear Programming Model

TL;DR

The paper tackles the challenge of EV charging infrastructure in Vietnam by formulating a Mixed-Integer Nonlinear Programming model to optimally place Level-2 and Level-3 charging stations in Ho Chi Minh City. The approach, solved with Gurobi using Brand-and-Cut, integrates seven cost components and leverages heterogeneous data (POIs and residential points) to meet demand while respecting a budget and driving-range constraints. Key contributions include a region-specific MINLP formulation, a detailed notational and constraint framework, and empirical results showing efficient convergence (MIP Gap $< 10^{-4}$) and actionable station distribution patterns that balance central and suburban needs. The study demonstrates practical implications for minimizing total costs and informing policy and operator strategies, with data and code available for replication and extension to broader regions.

Abstract

Vietnam is viewed as one of the promising markets for electric vehicles (EVs), especially automobiles, when it is predicted to reach 1 million in 2028 and 3.5 million in 2040. However, the lack of charging station infrastructure has hindered the growth rate of EVs in this country. This study aims to propose an optimization model using Mixed-Integer Nonlinear Programming to implement an optimal location strategy for EVs charging stations in Ho Chi Minh City. The problem is solved by Gurobi using the Brand-and-Cut method. There are two perspectives, including Charging Station Operators and EV users. In addition, 7 kinds of costs are considered. From 1509 Point of Interest and 199 residential areas, 134 POIs were chosen with 923 charging stations to fully satisfy the customer demand. Furthermore, the effectiveness of the proposed model is proved by a minor MIP Gap and running in a short time with full feasibility.

Figures (7)

• Figure 1: Compiles of main research directions about EV placement
• Figure 2: The branch is cut as a dotted line because of an infeasible solution. The normal line - branch continues to develop to achieve the best solution until dominated by the upper bound (UBD).
• Figure 3: Experiments schedule with the crawled dataset from Section. \ref{['sec:data']}, passed through Gurobi solver.
• Figure 4: Distribution of station locations (a) proposed (blue dots) and (b) existing (red dots) .
• Figure 5: $\#\text{station}_{\text{Level}-2}$ and $\#\text{station}_{\text{Level}-3}$ by Administrative Unit.