Wasteful base stations can be turned efficient via this model

‘A Predictive Model for Minimising Power Usage in Radio Access Networks’ by Emmet Carolan, Seamus C. McLoone, and Ronan Farell (Dept. Of Electronic Engineering, Maynooth University, Co. Kildare Ireland)

Best Paper award at MONAMI 20157th International Conference on Mobile Networks and Management

Not all the innovation is done via researching completely new designs. Rethinking the old, improving it and building on it can often be just as beneficial. In this vein, E. Carolan et al. suggest a neat way of using artificial neural network to optimize the power output of base stations.

Currently, the Radio Access Networks are largely designed in a way that they provide the peaking output all the time. This means a lot of power is used by them, or more specifically by cooling down the base stations (>50% of the overall power consumed). By powering down the base station during the time of lower demand, this excess energy could be cut down.

The determinants of level of connection demand are identified as time of day, day of the week, location, and special events. Several possible methods of designing the system are identified, and then the Artificial Neural Network is chosen as the most viable one, also based on suggestions from independent sources. Backward propagation of errors was chosen as the training method for this network. The whole process had three steps, creation of the network, training of the network, network’s prediction.

The team rightly identified that this model would save not only energy, but also a portion of unused radio spectrum, which is becoming a valuable commodity. The work adds to its real-life value by counting with the fact that turning off a base station does not happen immediately, but rather takes some time, so it needs to be planned well ahead. In the model, coverage area of each base station is divided into squares. It is working with two approaches to the base stations, either shutting them down completely, if they are not required, or letting them run on an energy saving scheme.

To see the results of their work, researchers have run three months of traffic data from 1145 base stations in Dublin county, Ireland, through the model. The graph displaying power saving ratio, and total traffic load over the time of the day, shows neatly when the most of the power can be saved.

The power saving ratio is the optimized power consumption divided by the the original power consumption.

The authors also ran similar models through the simulation, showing that their method significantly outperformed the other ones. In the conclusion they point out that this model was focused on energy saving, but a very similar design could be used to focus directly at tackling the radio spectrum limitation, primarily during off-peak hours.
Want to learn more? Get the full paper on EUDL.

Do not miss your chance to sign up for this year’s 8th EAI International Conference on Mobile Networks and Management. MONAMI 2016 will take place in Abu Dhabi, United Arab Emirates between October 23–25, 2016. Paper submissions are still open. We invite you to read the instructions for submitting a paper here.