Optimized one-way electric vehicle carsharing system in campus environment using heterogeneous non-dominated sorting genetic algorithm II

The general public are attracted to the carsharing systems based on the cost associated with the system and its efficiency in terms of service delivery. The accessibility and proximity of vehicle stations to its customers influence the service delivery; that is, the customer’s travel distance between origin and destination from the vehicle stations and the availability of vehicles at stations when the need arises determine the efficiency of the system. Conversely, the fleet size, station number and availability of vehicles at the designated locations at the right time determines carsharing system establishment and operation costs. Two main objectives are needed to optimize, the first one is the quality of service which is represented by the percentage of serving demands and the second one is the cost. Also, this optimization has to be built based on the stochastic nature of the environment. The previous works have concentrated on large scale type of environment where the demands can be equally distributed in most cases. However, small scale types of environments were almost ignored although they required customization when proposing a solution. In this study, the area that has been chosen is UPM campus, where the most crowded spots are easy to select because they are close to the faculties buildings. Next, the locations of the stations have to be placed in the environment based on the defined crowded spots. Once this is done, the simulation model can be built. The simulation model is responsible on generating the demands based in different times and locations. Each demand is generated in certain location named source of demand, at certain time, and it requires serving to another location called destination of demand. Our goal is to design the carsharing system which is combined of set of stations with certain size of each station and set of electrical vehicle some of them are parked while others are navigating to serve the demands. This design has to meet two aspects of satisfaction: one for the user and the other for the provider. This design leads to define the decision space, the objective function, and the constraints. MATLAB has been used for implementing our EV simulation and optimization. EV Carsharing measures that have been generated is the percentage of the number of unserved demand over the total number of generated demand with respect to time. NSGA-II heterogeneous has achieved better performance than the benchmark NSGA- II. NSGA-II heterogeneous has been compared with the benchmark NSGA-II from the perspective of multi-objective optimization. The evaluation results showed that NSGA-II heterogeneous has fulfilled more demands than NSGA-II homogenous with a percentage of 90% of the former comparing with 70% only for the latter. In addition to that, NSGA-II heterogeneous has achieved more than 70% hypervolume comparing with nearly 50% of hypervolume for NSGA-II homogenous for the case of non- equally distributed demands. This means more exploration power for NSGA-II heterogeneous which is supported by the higher number of non-dominated solutions.

Text FK 2019 153 - ir.pdf Download (1MB)

Actions (login required)

View Item