Proactive Route Planning for Electric Vehicles

TL;DR

The paper tackles Proactive EV Route Planning (PERP), addressing the challenge of routing for streams of EVs with partial recharging while reducing charging-station queues. It models the problem on a Graph with Time Dependent Self-loops (GTDS) and proves NP-hardness, then proposes a two-phase algorithm plus an influence-factor heuristic to select proactive charging paths. A forward-backward procedure computes all optimal charging paths (OCP) for each request, and influence evaluation guides path choice to minimize direct impact on future requests. Experimental results on real multi-city data show up to a 50% reduction in total travel time and substantial improvements in scalability and queue handling, especially with larger lookahead, making PERP practical for growing EV networks.

Abstract

Due to the limited driving range, inadequate charging facilities, and time-consuming recharging, the process of finding an optimal charging route for electric vehicles (EVs) differs from that of other vehicle types. The time and location of EV charging during a trip impact not only the individual EV's travel time but also the travel time of other EVs, due to the queuing that may arise at the charging station(s). This issue is at large seen as a significant constraint for uplifting EV sales in many countries. In this study, we present a novel Electric Vehicle Route Planning problem, which involves finding the fastest route with recharging for an EV routing request. We model the problem as a new graph problem and present that the problem is NP-hard. We propose a novel two-phase algorithm to traverse the graph to find the best possible charging route for each EV. We also introduce the notion of `influence factor' to propose heuristics to find the best possible route for an EV with the minimum travel time that avoids using charging stations and time to recharge at those stations which can lead to better travel time for other EVs. The results show that our method can decrease total travel time of the EVs by 50\% in comparison with the state-of-the-art on a real dataset, where the benefit of our approach is more significant as the number of EVs on the road increases.

Figures (8)

• Figure 1: The origin, destination, and driving range of two EVs, along with the locations of CSs are shown. If the green EV charges at CS1, then the travel time of the red EV gets extended, as the red EV has to wait until CS1 becomes available. If the green EV charges at CS2 instead of CS1, that results into a shorter travel time for the red EV. Both paths have the same travel time for the green EV.
• Figure 2: Example of a GTDS where each $v_i$ represents vertices, $c_1$ and $c_2$ represent two CSs. The weight $w(e)$ and resource unit $u(e)$ of each edge is shown as a pair of $<w,u>$ values with each edge. The shown table is the TRT, with a column for each self-loop for 8 timeslots. Each pair $<1,u>$ means that, by spending a timeslot, EV can gain $u$ unit of charge. If a timeslot is reserved at a CS, $u$ is zero for that entry (e.g., the first 3 timeslots in $c_2$).
• Figure 3: GTDS of Example 4. $l_1$ and $l_2$ are the source and destination, $c_1$ and $c_2$ are CSs with effective charging rate $2$ and $1$ per timeslot, respectively.
• Figure 4: Performance for varying the number of EVs: a) Total travel time comparison with the baseline in RDS b) Improved travel time by proactiveness in RDS c) Relative charge time in RDS d) Relative wait time in RDS e) Relative drive time in RDS f) Mean query time in SDS.
• Figure 5: Performance for varying the travel distance in SDS when the number of EVs is 1000: a) Total travel time in SDS b) Mean query time in SDS.

Theorems & Definitions (8)

• Definition 1
• Example 1: GTDS and TRT
• Example 2: Computing total time of a specific path
• Lemma 1
• proof
• Example 3
• Example 4
• Example 5