Neural networks for modelling the energy consumption of metro trains

Abstract

This paper presents the application of machine learning systems based on neural networks to model the energy consumption of electric metro trains, as a first step in a research project that aims to optimise the energy consumed for traction in the Metro Network of Valencia (Spain). An experimental dataset was gathered and used for training. Four input variables (train speed and acceleration, track slope and curvature) and one output variable (traction power) were considered. The fully trained neural network shows good agreement with the target data, with relative mean square error around 21%. Additional tests with independent datasets also give good results (relative mean square error = 16%). The neural network has been applied to five simple case studies to assess its performance – and has proven to correctly model basic consumption trends (e.g. the influence of the slope) – and to properly reproduce acceleration, holding and braking, although it tends to slightly underestimate the energy regenerated during braking. Overall, the neural network provides a consistent estimation of traction power and the global energy consumption of metro trains, and thus may be used as a modelling tool during further stages of research.

Predicting the traction power of metropolitan railway lines using different machine learning models

“…Speed and acceleration have been validated for MetroValencia itself in previous studies [26,54,62]. These results validate the veracity of our model and the possibility of applying different ML approaches to obtain better results than the most common ML model used: artificial neural networks (see Table 4).…”

“…Analysis of possible underfitting/overfitting problems After obtaining the best tuned model, it is important to analyse if it presents underfitting/overfitting. In a first instance, we used the 10-fold stratified cross-validation method to validate the ML models as we stated previously, this is complemented with the early-stopping method to prevent the ML models from overfitting [54]. Later, we implemented two strategies to analyse whether the selected model presents underfitting/overfitting: first, we implemented the learning curve method [55,56], a sampling method which monitors the increasing costs and performance of the model as larger amounts of data are used for training, and finds out when future costs outweigh future benefits; then, we implemented the grid search method to verify that the parameters obtained in the previous step do not overfit the tuned model.…”

Multi objective ant Colony Optimisation to obtain efficient metro speed profiles

“…Real energy consumption, travel time and speed profile data were measured along these lines to calibrate and validate the simulation model as well as to define the optimisation problem that constitutes the framework to compare MOACOr and NSGA-II. More details regarding data gathering and processing can be found in Martínez Fernández et al 29…”

“…This is the approach chosen for this study, which is based on the time-step simulator presented by Domínguez et al 35 but also incorporates a neural network to obtain the energy consumption. 29…”

Slab Track Optimization Using Metamodels to Improve Rail Construction Sustainability

“…Although many of them have been extensively compared in the past, deciding which one is better depends on the specific problem to be solved. That said, there are certain models that are more commonly used, such as polynomic regression, neural networks (García-Segura et al, 2018;Martínez Fernández et al, 2019a) or kriging (Penadés-Plà et al, 2020b). In this paper, kriging has been chosen as it is more flexible than polynomials and is less timeconsuming than neural networks (Simpson et al, 2001).…”