Energy-Efficient Timetabling Approach Considering Varying Train Loads and Realistic Speed Profiles

Zixuan Zhu, Fangsheng Wang, Rudong Yang, Zhibin Jiang, Ruihua Xu, Pieter VansteenwegenJournal of Transportation Engineering Part A Systems 2024 · DOI

TL;DR

Researchers have developed a new metro timetabling model that slashes energy consumption by up to 11.9% by better synchronizing braking and accelerating trains. Unlike previous models, this approach accounts for realistic speed profiles and fluctuating passenger loads, using an improved Adaptive Large Neighborhood Search (ALNS) algorithm to find the "sweet spot" for energy efficiency.

What Problem Does This Paper Address?

Metro systems are energy-hungry, with traction systems (the power used to move trains) accounting for about half of their total consumption. Regenerative braking—where a braking train converts kinetic energy back into electricity—is a common solution. However, there is a catch: in most systems, this energy must be used immediately by another nearby train that is accelerating. If no train is accelerating in the same power section, that energy is simply wasted as heat.

Existing research into "energy-efficient timetabling" often relies on oversimplified assumptions: they assume trains always have the same weight (ignoring passenger fluctuations), use idealized constant acceleration/deceleration rates, or stick to uniform headways (fixed time gaps between trains). This paper tackles the complexity of the real world by integrating varying train loads, realistic speed profiles, and bi-directional traffic into a single optimization model.

Methods

The authors formulated the problem as an integer energy-efficient timetabling model. The goal is to minimize total energy consumption by adjusting two main "levers":

  1. Dwell Times: How long a train stays at a station.
  2. Nonuniform Headways: Allowing the time interval between successive trains to vary slightly rather than being a fixed constant.

To solve this, they moved away from the commonly used Genetic Algorithms (GA), which can struggle with "premature convergence" (getting stuck on a decent but not great solution). Instead, they designed an Improved Adaptive Large Neighborhood Search (ALNS) algorithm.

The ALNS works by iteratively "destroying" and "repairing" parts of the timetable. The authors customized this by creating specific heuristics that randomly add/subtract time or swap values within the allowed ranges for dwell times and headways. The "adaptive" part means the algorithm tracks which heuristics are performing best and gives them a higher probability of being used in future iterations.

Datasets

The model was validated using practical data from Shanghai Metro Line 17:

  • Route: 13 stations across a bi-directional line.
  • Power Grid: Three distinct electricity supply sections.
  • Passenger Data: Historical sectional passenger volume data from November 2, 2019, split into 30-minute intervals to calculate varying train mass.
  • Operational Constraints: Realistic turnaround times at terminal stations and safety-mandated headway limits.

Key Insights and Ideas

  • Mass Matters: By using sectional passenger volumes, the model recognizes that a packed train leaving a city center requires significantly more energy to accelerate—and generates more energy when braking—than an empty one.
  • The Synchronization Puzzle: Energy efficiency isn't just about driving slower; it’s about "coincidence." The model shifts departure times by seconds to ensure that a train in Section A is braking exactly when another train in Section A is pulling out of a station.
  • Dwell Time vs. Headway: A surprising finding was that dwell time adjustments have a more significant impact on energy consumption than headway adjustments. Small tweaks to how long a train sits at a platform provide more flexibility for synchronization than changing the gap between trains.
  • ALNS Superiority: The improved ALNS was significantly more efficient than standard Genetic Algorithms, finding better solutions in the same amount of computation time (600 seconds).

Key Results

  • Energy Savings: The proposed approach achieved a maximum energy reduction of 11.9% compared to the current "real-world" timetable.
  • Algorithm Performance: In 360 test runs, the ALNS found high-quality solutions with a 94.2% success probability, consistently outperforming the baseline GA.
  • Nonuniform Advantage: The results proved that allowing headways to vary (nonuniform) is superior to keeping them fixed (uniform), as it provides the mathematical "room" needed to align braking and acceleration phases across multiple trains.

Limitations and Future Work

The authors acknowledge a primary trade-off: Energy efficiency vs. Passenger satisfaction. By increasing dwell times or headways to save energy, you might inadvertently increase passenger waiting times or crowding. Because of this, the authors note that this specific model is most suitable for off-peak hours when passenger demand is lower and slight delays are less disruptive.

Future Work: The authors suggest extending the model into a "multi-objective" optimization that balances operating costs (energy) directly against passenger service quality (travel time) to create a more holistic tool for metro operators.

Summary updated

Abstract

Energy saving has become a key concern of metro systems to reduce operating costs and meet the requirements of sustainable development. Energy-efficient timetabling, an effective way to reduce traction energy consumption, has gotten much attention in recent decades. However, limited research about energy-efficient timetabling simultaneously considered realistic speed profiles and the influence of train loads, which change greatly in different sections and significantly affect energy consumption. Therefore, this paper proposes an energy-efficient timetabling approach to improve regenerative energy utilization based on varying train loads and realistic speed profiles for bidirectional metro lines. First, based on the operational and safety constraints, this paper constructs an integer energy-efficient timetabling model where dwell times and nonuniform headways are taken as decision variables. Besides, the turnaround process is also considered to ensure rolling stock circulation efficiency. Sectional passenger volumes are employed to determine the varying train loads. Energy consumption is calculated based on realistic speed profiles. Owing to the complexity of the model, an improved adaptive large neighborhood search (ALNS) algorithm is designed to solve this model according to its characteristics. Finally, numerical examples are conducted based on the practical data from Shanghai Metro Line 17. The results indicate that the proposed approach effectively solves the energy-efficient timetabling problem with varying train loads and realistic speed profiles for bi-directional metro lines. The approach is suitable for off-peak hours and can reduce energy consumption effectively, and the reduction is up to 11.9% for the presented cases. The improved ALNS algorithm can find good solutions within reasonable times in different cases. Moreover, compared with the widely used genetic algorithm in the literature, the improved ALNS algorithm is more efficient in solving the proposed energy-efficient timetabling problem in different cases.
BibTeX
@article{zhu2024energy,
  author = {Zixuan Zhu and Fangsheng Wang and Rudong Yang and Zhibin Jiang and Ruihua Xu and Pieter Vansteenwegen},
  title = {{Energy-Efficient Timetabling Approach Considering Varying Train Loads and Realistic Speed Profiles}},
  journal = {Journal of Transportation Engineering Part A Systems},
  year = {2024},
  doi = {10.1061/jtepbs.teeng-8485},
}
Figure