Stockholm commuter train operator Stockholmståg is using a 'commuter prognosis' algorithm to forecast potential delays.

SWEDEN: Mathematician Wilhelm Landerholm of Queue AB and Stockholm commuter train operator Stockholmståg have developed an algorithm which can be used to forecast potential delays, helping network control staff to make decisions which will minimise the potential impact and also prevent ripple effects before they would have occurred.

Real-time feeds from across the network have been used to collect data on the actual arrival and departure time of every Stockholmståg train over a long period, and this is used to predict the impact of any disruption on future performance.

When a delay occurs, the mathematical model can be used to dynamically simulate the impact on every train to predict the time it will arrive at each station and the likely effects for up to 2 h ahead. Controllers can simulate various options for responding to the delay, helping optimise decision making.

The model can also used for passenger information, with a smartphone app to be launched this year to provide passengers with advance warning of likely delays.

‘We have built a prediction model using big data, that lets us visualise the entire commuter train system 2 h 2 two 2 h 2into the future,’ explained Mikael Lindskog of Stockholmståg. ‘We can now forecast disruptions in our service and our traffic control centre can prevent the ripple effects that actually cause most delays.

‘Today the traffic control centre analyses the delays manually in order to prevent future delays,’ he added. ‘By automating the forecasting we can raise our service level significantly. The “commuter prognosis” will be the first automated forecasting model of its kind. In the long term it is possible that this will change how traffic control centres work all over the world.’

Possible extensions include incorporating real-time data on passenger loadings to gain a better understanding of performance.

Landerholm told Railway Gazette International the model could be applicable to any rail network with timetabled services where detailed real-time data is available.