AI for Traffic planning of busses

Important results from the project

The purpose of developing the possibilities for Portal+ to deliver AI-based solutions has been fulfilled. It is now possible to train models based on neural networks. Sub-goals 1, 2 and 4 are met according to the business´ wishes and inputs. The focus was on getting a model better instead of two similar ones as the business did not realize the total value in sub-goal 3´s model.

Expected long term effects

The project has delivered a model for optimizing fuel consumption during transportation within the BPL system. This has resulted in a plan that is 5% more fuel efficient. As stated, a bus at Byberg & Nordin runs an average of 7000 km per year, and draws an average of 3 liters per 10 kilometers. On the 200 vehicles this project has been carried out on, it results in an average saving of 1050 liters of fuel per bus and year. A total of about 210,000 liters of fuel. Converted to CO2 and calculated on diesel, it gives according to WTW climate impact a saving of about 546900 kg CO2 per year in emissions.

Approach and implementation

The solution was implemented in several steps. First, a model was developed based on neural network technology that can predict fuel consumption for a specific transport based on a large number of input parameters. The model was trained on the amount of data we had available and then used to simulate the lowest total consumption for an appropriate amount of transports. The model gives suggestions on which vehicles should be planned on which transportation to provide the lowest consumption.