Data Centers Reliance On Continuous Power – In a bid to reduce emissions, renewable sources are being used where possible – but this creates new problems for the grid, making it harder to match generation and consumption. Data centers could help to create a balance, through techniques referred to as “demand response” but so far it’s proven difficult to enlist their help.
All the world’s economies are attempting to reduce carbon emissions by increasing the share of renewable sources in their electricity generation, and reducing that provided by fossil fuels. However, there are two problems with this.
Fluctuating power
Firstly, apart from hydroelectric power, renewables are mostly intermittent. Solar panels and wind turbines only deliver energy when the sun shines or the wind blows, and can’t be switched on as required.
And secondly, the fossil fuel-powered capacity that is being retired is exactly the steady, readily available capacity that the grid needs, providing a continuous baseload, and also extra flexible capacity as needed.
The electricity grid has to satisfy a fluctuating demand and there are two big factors where long term policies designed to reduce emissions could actually add to the burden on the grid.
Electricity is being proposed as a replacement for fossil fuels in cars and heating. But this will increase the demand for electricity and it only reduces emissions if green electricity can be increased to match that demand.
RELATED: AMD INSIDE INTEL OUTSIDE SAYS CLOUDFLARE
To respond to changing supply and demand, the grid has to become flexible. According to Mohan Gandhi, a civil engineer and analyst at the New Bridge Founders thinktank: “As intermittent renewables penetrate further into the generation mix, flexibility becomes an increasingly important feature of the electricity system.”
Some of this flexibility is based on moving electricity to where it is needed, but transmitting power is costly and involves losses and the cables may not even be there: “Renewables are actually being built faster than cables can be laid,” says Gandhi.
“In Germany, wind generation in the north has grown enormously, but the interconnection cables between the north and south are yet to be built.”
Instead of moving electricity around, another approach is to shift demand towards times when electricity is cheaper or more available – an approach dubbed “demand response.”
This can be as simple as offering consumers a cheaper tariff for night-time electricity (in the UK often referred to as Economy 7).
In industry, energy use is more concentrated, and there is potential for more advanced methods including on-site generation and stored energy, so industrial sites can temporarily shift their load completely off the grid, or even become an energy source, feeding power into the grid.
“Demand response is often the most economical form of flexibility because it requires few new transmission or distribution investments,” says Gandhi. It also has a lot of potential: The European Commission estimates that Europe as a whole could deliver 100GW of demand response power (and this is expected to rise to 160GW by 2030).
However, the European grid is currently only accessing around 20GW of available demand response capacity. Globally, Gandhi estimates that 20 percent of the world’s electricity consumption will be eligible for demand-response by 2040.
As a sophisticated and significant electricity user, believed to be using around one percent of the US grid’s output, digital infrastructure can play a big role here. “Data centers, with their real-time management and workload flexibility, are good candidates for demand response schemes,” says Gandhi.
“They can ‘shift’ load outside peak hours, or deliver surplus energy stored in their batteries and on-site generators to the grid at times of undersupply.”
It’s been suggested that a group of data centers could help shift demand by migrating their loads amongst themselves to make use of the cheapest and greenest electricity at a given moment. “Many typical data center workloads are delay-tolerant, and could be rescheduled to off-peak hours,” says Gandhi.
There are drawbacks to this. Firstly, if a data center is running profitable workloads, then it costs money to move them elsewhere, and the most cost-effective use of that resource is to run it at capacity as long as possible. And secondly, the customer who owns the data may need to ensure that is processed in a given location to comply with local regulations.
Reducing power consumption
It’s actually possible for data centers to reduce their power demands without affecting IT workloads. Research by the Lawrence Berkeley National Laboratory (LBNL) found that energy consumption could be reduced by five percent in five minutes, and 10 percent in 15 minutes by making changes such as setting a temporarily higher air temperature.
Beyond this, demand response approaches tend to use the facility’s uninterruptible power supply (UPS). This is designed to support the data center when the grid fails: there is an alternative source of power (usually diesel gensets), and some energy storage (typically batteries) that will support the data center while the local power starts up.
Why not use the batteries, or even switch to diesel for a few hours, when energy is expensive? “When power is expensive, you can use energy from batteries, not the grid,” says Janne Paananen, technology manager of energy equipment company Eaton. “This gives savings in cost of energy. You can do it yourself.”
Beyond the DIY approach, there are systems managed by the utilities, which work in a surprisingly simple way. The grid frequency in the UK is 50Hz (plus or minus one percent), but it varies at heavy loads. The utilities use this to regulate the power – the grid detects the change in frequency and uses that to switch on extra capacity.
Because there are industries with their own generating capacity for backup, the electricity industry has come up with a scheme called Firm Frequency Response (FFR), in which those third party resources are turned on in response to the same change in frequency.
Data center UPS systems are designed to switch on immediately, and can be hooked into this sort of scheme.
FFR is in operation in Ireland and likely to come onstream in the UK shortly. Eaton is working with the FFR scheme in Ireland, says Paananen. ”With fast frequency response, you are providing services for the grid, and getting revenue by providing those services. Instead of responding to the cost of energy, you respond to a real-time signal.”
In Ireland, facilities on the FFR program get a signal roughly once a month, says Paananen. “Normally the frequency deviation lasts for only a few seconds.” This is a level of usage that traditional lead-acid batteries can readily support – and if FFR takes the place of a scheduled battery test, it can actually create less stress to the system.
These systems are proven, says Paananen. On 8 August 2019 in the UK, two power plants went down, the frequency of the grid changed, and that signaled various responses, so numerous factories and facilities went off-grid.
All this has been possible for years, but – as with any new idea – the big hurdles are making it pay, and gaining users’ trust. Utilities are prepared to offer cheaper electricity at different times, and even pay consumers to take themselves off-grid. But will data centers take them up on this?
Back in 2013, Ciaran Flanagan of ABB told DCD: “Demand response programs (DRPs) have not only become a tool for grid operators to manage demand, but also a source of revenue for DRP participants.
DRPs are in operation today in many commercial and industrial sectors but, ironically, data centers are largely non-participants, even though they are the fastest-growing part of the grid’s load.”
A big reason for data centers’ reluctance is that they make profits from continuous availability, and sharing support systems might increase the risk of failure.
“Participating in demand response programs may reduce the availability or lead to a higher risk of downtime. This risk is exacerbated by the potential surrendering of control to aggregators,” says Gandhi. “Data centers are typically in the business of avoiding downtime, minimizing risk and maximizing availability.”
