The design of sustainable mobile networks is key to reduce their impact on the environment, and to diminish their operating cost. As a solution to this, we advocate Energy Harvesting (EH) Base Stations (BSs) that collect energy from the environment, use it to serve the local traffic and/or store it in a battery for later use. Moreover, whenever the amount of energy harvested is insufficient to serve their traffic load, BSs purchase energy from the power grid. Within this setup, a smart energy management strategy is devised with the goal of diminishing the cost incurred in the energy purchases. This is achieved by intelligently controlling the amount of energy that BSs buy from the electrical grid over time, by accounting for the harvested energy, the traffic load, and hourly energy prices. The proposed optimization framework combines pattern forecasting and adaptive control. In a first stage, harvested energy and traffic load processes are modeled through a Long Short-Term Memory (LSTM) neural network, allowing each BS to independently predict future energy and load patterns. LSTM-based forecasts are then fed into an adaptive control block, where foresighted optimization is performed using Model Predictive Control (MPC). Numerical results, obtained with real-world energy and load signals, show cost savings close to 20% and reductions in the amount of energy purchased from the electrical grid of about 24%, with respect to a heuristic scheme where future system states are not taken into account.

Smart Energy Policies for Sustainable Mobile Networks via Forecasting and Adaptive Control

Gambin, Angel Fernandez
Investigation
;
Rossi, Michele
Supervision
2018

Abstract

The design of sustainable mobile networks is key to reduce their impact on the environment, and to diminish their operating cost. As a solution to this, we advocate Energy Harvesting (EH) Base Stations (BSs) that collect energy from the environment, use it to serve the local traffic and/or store it in a battery for later use. Moreover, whenever the amount of energy harvested is insufficient to serve their traffic load, BSs purchase energy from the power grid. Within this setup, a smart energy management strategy is devised with the goal of diminishing the cost incurred in the energy purchases. This is achieved by intelligently controlling the amount of energy that BSs buy from the electrical grid over time, by accounting for the harvested energy, the traffic load, and hourly energy prices. The proposed optimization framework combines pattern forecasting and adaptive control. In a first stage, harvested energy and traffic load processes are modeled through a Long Short-Term Memory (LSTM) neural network, allowing each BS to independently predict future energy and load patterns. LSTM-based forecasts are then fed into an adaptive control block, where foresighted optimization is performed using Model Predictive Control (MPC). Numerical results, obtained with real-world energy and load signals, show cost savings close to 20% and reductions in the amount of energy purchased from the electrical grid of about 24%, with respect to a heuristic scheme where future system states are not taken into account.
2018
2018 IEEE Globecom Workshops, GC Wkshps 2018 - Proceedings
9781538649206
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11577/3296904
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 0
social impact