Introducing the GEYSER Optimization Strategies Simulator
|Dr. Tudor Cioara, is an active member Distributed Systems Research Laboratory of the Technical University of Cluj-Napoca. In 2012 obtained his Ph.D. in Computer Science with the thesis "Context aware adaptive systems with applicability in green service centers." Research areas in which he is currently working refers to Context Awareness, Green IT, Ambient Assistant Living and Artificial Intelligence.|
GEYSER aims to contribute to the on-going efforts of integrating urban data centres to their local smart grids and cities. As such within the project we view data centres as active energy load resources in urban environments that can leverage on the local green energy and ancillary services markets.
To that end, we have been working on the GEYSER Optimization Engine: a tool that helps data centres exploit their high demand flexibility to provide an optimal capacity and operational planning; in doing so data centres can shape their power consumption profile to meet various objectives in collaboration with requests coming from the local grid and thus allowing them to participate in demand response programs.
We have defined three optimization alternatives working at different time scales: day-ahead, intra-day, and near-real time. The day-ahead operation planning estimates the data centre energy to be contracted from the day-ahead market for the next operational day; it accordingly plans the actions for data centre operation so as to meet the contracted levels for following day. Such actions are constrained by considering the contracted energy supply levels and the energy consumption and production predictions. During the next operational day, the execution of the optimization action plan is in turn tracked along with the degree of prediction conformity by using two levels of granularity: intra-day planning (few hours ahead) and near real-time (few minutes ahead). Due to the inherent inaccuracy of the predicted data, errors may appear and as such corrective actions must be considered for the next few hours or for the next few minutes. Such corrective actions may include, but not limited to, buying energy from the intra-day marketplace or from the near real-time marketplace, respectively. Naturally, the closer we get to the real time, the highest the cost of implementing the corrective actions.To shape and optimize the DC’s power consumption profile, we have defined and used flexibility mechanisms associated to energy consumption of the following systems:
- Electrical Cooling (i.e. dynamically using non-electrical cooling mechanism such as thermal storage);
- IT Computing Resources (i.e. by shifting delay tolerant workload at different time slot);
- Energy Storage (i.e. charging/discharging the batteries); and
- Energy Generation (i.e. planning the periodic maintenance of the diesel generators when there are energy production deficits or in the case of high energy prices or to avoid coincident energy demand peaks).
To prove the effectiveness of the GEYSER optimization methodology and defined flexibility mechanisms, we have implemented a simulation engine which can be accessed at:
In this first version we have implemented the following optimization policies:
- maximize the usage of renewable energy produced on-site;
- shift DC’s energy demand to achieve a closer match to the energy profile contracted from the day-ahead marketplace;
- maximize the usage of renewable energy produced in smart grid;
- optimize the DC energy-related profit; and
- minimize the DC operational costs.
Feel free to check the first prototype for yourself!
After registering, you may load any of the pre-defined scenarios or define a new one. Each scenario provides the optimization engine with an initial data centre energy consumption profile, data describing its operation over time, and a GEYSER defined optimization policy. You can then view the data centre‘s current and forecasted energy budget along with the optimization engine’s decisions on the day-ahead, intra-day, and near real-time action plans.
We will of course continue to develop and further refine our optimization engine; meanwhile should you have any questions, comments, or feedback feel free to contact us!