Optimisation of a Commuter Train System’s Energy Consumption: : A Statistical Approach

Sammanfattning: In recent years the number of passengers travelling by train have increased and so has the requirements for punctuality and energy efficiency. The performance of the requirements depends on how the driver operates the train which in turn depends on the drivers skills and experiences. A Driver Advisory System(DAS) that gives the drivers guiding can increase the operational performance on these requirements.  The aim of this thesis project was to investigate if the available data partly containing information about the trains velocity, acceleration and effect usage can be used to develop a system that aids the train drivers to reach the next station in time whilst minimising the consumed energy. The project was di-vided into two parts. In the first part different regression models and data setups were tested to see how well they could capture the effect usage. The tested regression techniques were linear regression and support vector regression, both techniques giving a good result with aR2over0.9. From the tried data setups the results shows that one estimated model could be used for all the trips in the system and the trips could be looked at as 1 or 3 phases. The second part of the project was to see if the estimated regression model could be used in an optimisation problem to find the best speed curve between two sta-tions. The results from the optimisation problem presents a solution between two stations with a lower energy consumption than the average historical trip. The optimisation results gives directives for the optimal way of driving the train as well, where the acceleration should not be over 0.85m/s^2 and where the max speed during a trip should not be higher than needed.