Smart Grid with Smart Data Centers (DEAN)
Agent-based intelligent reliable energy infrastructure to support growing market of large consumers (data centers, etc.).
Aim of the project
Investigate smart ways to manage power grid to achieve better power efficiency and reliability within the region. The energy can be optimally used when large consumers (including aggregated load) coordinate their power consumption among each other, generators and network operators. This also will improve power reliability, as network operators will be able to negotiate strategies to prevent overloads and faults or to recover network.
Data center can be thought as an industrial process, requiring large amount of energy, precise technologies and methodologies. Data center is a large scale facility for computing, that produces enormous amount of heat. There are many processes that require careful design for efficient, profitable and safe operation.
The project is about investigating ways of managing a power grid to achieve a better power quality, energy efficiency and better reliability. Can data centers participate in dynamic power balancing and “help” the network operators to manage the grid? We think this grid will be managed by collective effort of distributed software agents. We might find the ways to manage energy more efficiently, and reduce power losses and environmental impact of fossil fuel energy generation.
Data centers consume energy at the industrial scale, with high demands for availability. This motivates installations of power redundancy with large capacity, such as UPS, backup generators and chillers. High power bills and power consumption of data center prompts that both data center and power utility can benefit from demand response programs: peak shaving, load shedding, interruptible power contracts and capacity reserves. Data center can employ various methods of demand side management: exploit inertia of the cooling and heating systems, control pumps and chillers, shift IT load in space and time, move to self-powered mode (UPS, backup generator), store energy for later use (battery, chillers, heaters and etc) and others. Such data center as an active load can result in more efficient and sustainable power grid.
Why:
As an active load data center can participate in the load and power balancing within the power grid, rather than being a growing passive load on the network. The aim of this work is to facilitate energy efficiency of data centers within the regional power grid:
- Sustainable data center growth (several in the region);
- Cost savings (total energy bill);
- Flexibility for the grid – stability of the network;
- Alternative design – Solid State Transformer (SST) – as an enabler of an active load.
Method:
We want to apply agent based approach, modeling a data center, residential load, industrial consumer and utility as individual agents with their own goals. Initially the goals will be simple:
- Residential load – minimizing power bill
- Industrial consumer – ensure availability of the production process (not much regard to the power bill)
- Data center – minimizing power bill, ensure availability of the service, max participation in demand response programs
- Utility – maintain network stability (frequency regulation), smooth out peaks and dips of the power consumption, max utilization of energy in the network.
The resulting multi-agent system can be realized as cooperating agents; competing agents; selfish agents; or as a resource allocation problem (auctions).
The agents can be motivated by real-time prices, rewards for participating in demand response programs, time deadlines of various tasks, operational requirements (e.g. availability), and their own goals.
The resulting multi-agent system can optimize total energy use in the power grid while maintain its stability (frequency).
In order to provide such flexibility and independence for the data center, microgrid has to be integrated with the local energy network, however mainly operate in islanding mode.
In order to achieve this, it is necessary to resolve technical challenges of 1) grid and microgrid stability during transient periods; 2) dynamically changing grid topology; 3) self-healing of the grid and microgrid (fast protection and control).
As a control center for a microgrid, data center needs to 1) monitor real time weather and energy price information; 2) solve in real-time EDP problem; 3) perform selected demand response actions; 4) participate in local energy market; 5) negotiate and coordinate actions with local power network operators; and 6) have full control over its infrastructure and microgrid.
Updated: