Goal of the DESI project is to optimize the energy consumption of ICT production, considering both the domains of "ICT" and "Energy" in a unifying way. The operation of ICT components over the whole supply chain, up to customer networks, has to be taken into account and controlled with respect to varying traffic. This leads to new requirements for the power supply of widespread telecommunication networks.
Taking a unifying view on energy and ICT network makes it necessary to consider complex relationships between various restrictions on different systems. We use state-of-the-art methods of mathematical optimization to deal with this challenge. This includes a model-based description of network topologies and structures, the study of the net-wide influences of components under varying load, as well as the design and implementation of algorithms for the energy-efficient control of ICT production.
Questions from both domains Telecommunication networks and Energy supply have been studied at ZIB.
Load-adaptive telecommunication networks
Traffic demands in telecommunication networks heavily fluctuate in daily and weekly cycles.
(Source: European research network GEANT)
Modern load-adaptive networks can take advantage of these fluctuations by switching off parts of the network at times of low traffic and thereby saving energy. Various network configurations are visualized below: at peak traffic hours (center) and at low traffic hours (right) in comparison to a load-independent network (left).
The problem of finding energy-optimal network configurations is solved using mixed integer programs.
In practice, frequent reconfiguration of the whole network is not desirable. Therefore, another challenge is to find few (e.g. 3) scenarios over a certain time (e.g. one day) such that the overall energy-consumption is as low as possible. Solution approaches for this problem make use of the above models for the optimization of network configurations.
Optimal battery control
Batteries that are available at locations within the network of our project partner Deutsche Telekom can be used to store energy at suitable times. At a later time, the network components can then be supplied with power independently of the power grid.
The mathematical models are non-linear due to the physical properties of batteries. To solve these models we use SCIP, implementing various approaches.
By default, battery control is optimized with respect to dynamic energy prices; however, using other information is conceivable, such as availability or sustainability of energy production. This opens the chance to reduce fluctuations in the power grid and increase its stability, a feature that will become more and more important with the extended use of alternative and more volatile forms of energy.
