Data centres and the power system: demand, location and flexibility

Data centres and the power system: demand, location and flexibility

With large swathes of data required by the energy industry to help keep abreast of machine health and maintenance programs, as well as energy finance forecasting, data centres are required to house this Big Data. Large data centres are reshaping electricity systems, thanks to fast-growing loads, strict uptime needs and flexible demand potential, and their integration is changing how grids plan capacity, site infrastructure and balance supply with variable renewables. But there are challenges ahead: some constraints exist around which sites can reasonably accommodate these large data centres. From a practical perspective, some development will need to take place regarding how data centres can be integrated into wider information networks to establish data centres as flexible, low-carbon consumers.

One thing is certain: data centre growth will impact power systems and how planners, utilities, operators, and corporate buyers can manage demand, location, and flexibility without increasing costs or emissions. We examine the factors that influence and affect their integration.

Why do data centres matter for power systems?

Data centres are key to allowing the grid to grow in terms of size and complexity.

Scale and growth of demand, including AI and high density compute

As more complex technologies are folded into the grid, the demand for data centres in the energy industry is rising. To be useful as an analytical tool, AI requires a large amount of machine health data to forecast factors such as demand, supply, expected pricing and machine behaviour. It can then model future scenarios that predict how a plant, consumer or market may react in hypothetical scenarios, allowing market participants to make decisions, set prices or schedule maintenance. But this Big Data needs somewhere to live, and this is where data centres come in. 

Load shape basics 

Data centres themselves are huge machines that also require energy to power them. Their own base loads in themselves could result in spikes from batch data processing jobs. This is why it's important that data centres are holistically thought of in terms of assisting the grid, but also being fed by the grid. 

Uptime, latency and cooling 

Uptime, latency and cooling make these loads operationally unique because the management of infrastructure becomes more complex as well as more expensive. Elements like disaster recovery must be introduced to deal with high levels of uptime, while latency factors are addressed by designing specialised networks to accommodate them. Because data centres generate so much heat, cooling methods such as liquid cooling are required. 

Data centre growth, local prices and capacity planning

Links between data centres and centre growth, local prices and capacity planning are close. The growth of a data centre affects many areas, including planning from both a spatial and capacity level, as well as an infrastructure perspective. It can also increase costs and affect local economics through raised costs and increased economic activity in the local area. It’s not all negative, though - they can, however, bring employment and social benefits to the community. 

Measuring demand today and tomorrow

Knowing the efficiency of a data centre is paramount to its successful running, and there are various methods we can use to understand this, for example, by estimating the site and portfolio load of data centres' power versus total facility power. They enable site owners to assess the efficiency of their infrastructure. 

Lower is better when it comes to PUE (Power Usage Effectiveness) and WUE (Water Usage Effectiveness), both of which are measures of efficiency. As AI becomes more demand-driven, high-performance servers will require more power from data centres, which will call for more advanced cooling, including electrified cooling. 

Integrating renewables and low carbon supply

More demand for data centres will require more power, and this could represent a real opportunity for renewable energy providers, including smaller, decentralised grids. Off-site PPAs could be utilised, with consumers operating on a pay-as-produced model with 24-by-7 matching. These could be bolstered by guarantees of origin, allowing adherence to disclosure rules that buyers must manage. However, on-site solar and storage might face practicality issues, with space limits in urban sites. The role of interconnectors and market coupling is important in ensuring clean supply access: they link up different types of renewable energy across different grid regions, helping to increase investment in green technologies. 

Flexibility toolkit for operators and utilities

Operators and utilities can benefit from demand response levers, such as workload shifting or non-critical load shedding, which involves diverting power from non-essential services during high grid demand. 

Cooling set points alleviate the pressure on air conditioning units, allowing data centres to reach a certain temperature before air conditioning kicks in. Ice storage can be used for peak shaving by forming ice at night when demand is lower, an action known as peak shaving. Data centres can be even more flexible in terms of energy consumption by tracking carbon pricing and scheduling batch jobs and training windows to coincide with periods of lower demand.

Battery systems can be used to bolster the grid when data centre energy demand is high, creating reserves when energy supply is low but demand is high. Control architecture data, forecasting and automation can also be integrated for control and reliable flexibility of the energy required and consumed by data centres.

Backup generation, emissions and compliance

Backup generation will play a key role in data entry integration, utilising both renewable sources such as hydrogen fuel cell batteries, as well as fossil fuels. Diesel and gas will still play backup roles, but will be limited to run hour limits and have to adhere to strict air quality rules, with the regular testing of reported emissions and local air quality. This could lead to pathways to lower emissions using carbon capture and other emerging methods. 

Cooling, water and heat reuse

Large data centres require substantial amounts of cooling, resulting in significant water usage. This could become a problem due to water constraints and will require drought planning, including alternatives for water-stressed regions. Cooling technologies such as air cooling, which uses airflow to cool during heating, while liquid cooling requires a coolant and immersion cooling uses a dielectric compound to cool equipment. Heat can also be reused rather than discarded: heat reuse can be a key source of warmth for district heating.

Grid and market impacts

As the need for extra energy, resources and infrastructure increases with the wider integration of data centres, local network reinforcement needs could also increase, and issues could arise when deciding who pays for the upgrade, running and maintenance of these adjustments. 

This could potentially have a knock-on effect on wholesale and balancing markets, with price volatility occurring near clusters and mitigation strategies required to address disruptive price fluctuations. 

Procurement playbook for energy and capacity

One thing is for sure, the energy industry must prepare for the overarching effect the rollout of huge data centres will have on the supply and demand of power, and try to mitigate the negative impact on the wider grid servicing them. Contracting options will need to be structured specifically for retail, with wholesale access addresses differently, with specialist capacity charges and network tariffs developed for individual markets. Financial tools are available to help mitigate the impact, with hedging toolkits focusing on forwards and options posed as insurance against price spikes and increased risk. 

Success means smart siting, efficient cooling, clean power and flexible operations so data centres support the grid, cut emissions and control costs as demand grows.