AI is beginning to impact nearly every part of the energy sector, both upstream and downstream, particularly focused on the integration of more variable power generation sources, with increasing needs for accurate forecasting and load management, along with progress on decarbonisation.

‘As we move into the Fourth Industrial Revolution, grid operators, developers and consumers are harnessing AI, paving the path for a smooth transition to greater use of renewables,’ explains Thierry Martier, EY Global Digital and Innovation Lead for Energy, in a report on Why AI is a game-changer for renewable energy.

‘Better prediction is enabling improved demand forecasting and asset management, while AI’s automation capability is driving operational excellence, competitive advantage and cost savings,’ he notes. Advances in machine learning go hand in hand with emerging technologies, such as the internet of things (IoT), use of sensors and big data – offering the ability to handle complex tasks at speed.

Predicting grid capacity levels is paramount for a stable and efficient grid. ‘As renewables take up an increasing share of the grid, there is a loss of baseload generation from traditional fuel sources such as coal, which provide grid inertia via steam and gas turbines. Without grid inertia, power networks are susceptible to blackouts because of the intermittent nature of renewables. However, using sensor technology, solar and wind generation can provide a vast amount of real-time data, which allows AI to predict capacity levels,’ explains Martier.

‘AI’s automation capability is driving operational excellence, competitive advantage and cost savings.’ – Thierry Martier, EY Global Digital and Innovation Lead for Energy

Better weather forecasting 
AI also offers more reliable weather forecasting, like the IBM AI programme used by the US Department of Energy’s Sunshot Initiative. This combines self-learning weather models, historical weather data and real-time data with sensor networks and information derived by satellite imagery.

IBM has also created an AI-based platform in partnership with Omega Energia, a Brazilian renewable energy company, to improve renewable energy generation forecasting using geospatial and meteorological data analysis in the cloud. The powerful IBM PAIRS geospatial and analytics platform is claimed to have produced a 30–40% improvement in wind and solar forecast accuracy.

Improved trading and planning 
Big data sets also provide improved forecasts of power consumption during peak periods, helping to optimise bidding in wholesale power and balancing markets.

For example, the Origami energy data platform offers an AI-based software-as-a-service (SaaS) for energy traders to rebalance their position when addressing fluctuating renewable energy markets, using more granular and diverse data than historic data methods.

James Kelloway, Energy Intelligence Manager at the National Grid ESO, also recognises the value of more accurate predictions with respect to renewables. ‘Using AI we can control other power plants more accurately, like coal plants which take many hours to ramp up.’ What’s more: ‘Optimisation of system configuration can make renewables not only greener but potentially cheaper,’ he adds.

AI can also be used beyond central planning to play a role with machine-to-machine communication on the edge of the grid, improving reliability and combating grid congestion in complex, decentralised systems.

Demand forecasting 
AI is also useful for consumer demand forecasting in concert with smart meters. AI systems can predict a building’s thermal energy demand to produce heating and cooling at optimal times, in concert with home solar and battery systems.

Josh Lehman, Senior Director of Product Management at US energy storage firm Stem, estimates that AI-driven software has improved customer savings by about 5% year-on-year.

Battery storage optimisation 
AI is playing a pivotal role in providing demand flexibility using battery storage, by considering forecast demand, renewable energy generation prices, network congestion and other variables, to minimise the back-up energy needed from diesel generators, coal-fired plants or gas-fired peaker plants.

US-based SaaS platform provider AMS uses AI in battery storage systems to identify opportunities to purchase electricity from the grid when prices are low, and then sells back to the market when prices are high. Australia’s 150 MW Hornsdale battery operates an autobidder AI algorithm developed by Tesla that is claimed to have captured revenue streams several times higher than a conventional energy trader.

Improved operations and maintenance 
AI can also assist with operations and maintenance management, identifying exception incidents for predictive maintenance. The algorithms detect disturbances in real time of mechanical failure, improving power system reliability and efficiency.

‘Unexpected disruptions across the power industry can cost 3–8% of capacity and $10bn annual lost-production cost,’ says Brian Case, Chief Digital Officer at GE Renewable Energy. GE’s Predix software is embedded with AI-based algorithms to interpret industrial data for predictions on machine health as well as recommending actions to improve the efficiency of renewable energy assets such as wind farms.

Monitoring wind farms 
DNV provides an integrated approach to assure AI-enabled systems by assessing the development process using AI testing technologies. The company recently acquired Proxima Solutions, a Berlin-based AI-enabled SaaS provider for remote monitoring asset management of wind farms. Proxima’s Windlog and Energylog software tools will be merged with DNV’s GreenPowerMonitor business offering advanced analytics, real-time updates, asset management and predictive diagnostics to renewable asset owners.

DNV offers a suite of AI technologies including Corrosion.ai which provides AI and machine learning for inspection and monitoring of corrosion on ships. Its Battery.ai uses AI and empirical models for monitoring and verification of battery health. And NDT.ai provides automatic flow detection for weld radiographic (x-ray) tests during non-destructive testing.

The Norwegian assurance company also has an AI research centre located in Shanghai, China.

Case study: Transforming building energy consumption  
IBM Consulting’s Flex Platform uses AI, machine learning and IoT to balance a building’s green energy consumption more flexibly. ‘Reduction in energy consumption can help municipalities reduce their carbon footprint by adopting a flexible approach to power consumption,’ explains Phil Spring, EMEA Energy & Resources Leader, IBM Consulting.

The Flex Platform has been developed by IBM in partnership with Andel Energi, Denmark’s largest energy company, to handle fluctuating renewable energy sources. The European supermarket chain Salling Group is now starting to utilise the platform to balance consumption of green electricity in the grid, helping to advance decarbonisation. The group’s Netto supermarket stores in Denmark can now balance their power consumption by integrating electrical systems for heating, ventilation and cooling in on one building management platform.

The Flex Platform uses IoT sensors, AI, blockchain and the cloud to integrate energy aggregators and their customers with the energy system for real-time, intelligent grid optimisation. The AI-based platform balances the grid when renewables fluctuate or demand peaks, by analysis of connected assets such as heating, ventilation and air conditioning (HVAC) systems, water pumps and data centres, for optimal performance without reliance on reserve power plants.

Renewable energy sources are intermittent by nature, so balanced grid management requires new technologies like AI, IoT and blockchain
Photo: IBM
 

Spring recognises the complexity of the renewable energy ‘which brings together many ecosystems’. Although he sees a lot of focus upstream on power generation sources such as wind farms on- and offshore, he has particular interest downstream where energy is delivered to transport, cities and individual buildings, using a very distributed and decentralised energy system.

‘You need to balance many different parties if you’ve got solar PV and wind farms as a primary electricity generation source which is inherently intermittent – with peaks and troughs,’ he comments.

Typically, in western economies, two or three times as much energy will be delivered through the electricity system as before. ‘This level of activity calls for a smart grid to deal with electrification and intermittent supply, to maintain security of supply faced with the energy trilemma that people talk about. That’s what the Andel marketplace does,’ says Spring.

IBM started working on the project with the city of Copenhagen five years ago, connecting small-scale energy consumption assets like buildings into a smart grid, working with Danish grid operator Energinet. ‘The big issue was having all the wind farms in the east and consumption in the west – managing that intermittency and demand, balancing the network without creating constraints,’ he says.

The key was to create a flexible platform bringing together a variety of embedded technologies, like AI, machine learning and IoT, to model the energy attributes of a building, in order to forecast where flexibility is available for diverse connected assets.

Spring emphasises the need to be patient with this type of development. ‘The first lesson is: it takes time. It took five years to get the programme off the ground. Second, you need to think about these problems as ecosystems that require collaboration – whether you are a building owner, grid operator, aggregator or whoever – to create a new energy marketplace to solve the flexibility problem from an energy point of view.’

There is also the issue of regulation. ‘This also takes time because there are rules around what you can and cannot do when it comes to flexible power markets,’ he notes.

The plan is to roll out the Flex Platform to about 500 small Danish supermarkets, then to different types of commercial buildings, warehouses and schools, with a ‘common level’ of building management systems that can be hooked into the cloud as part of a flexible energy system.

Discussions are underway to extend roll-out of the Flex Platform to Germany and the UK. ‘The UK is considered to be quite progressive when it comes to these sorts of energy ecosystems. But it’s not just about getting energy companies onboard and deciding who will host the platform, but talking to many real estate and building owners to help them understand the business case in different markets extending through to electric vehicles. You have to look beyond the boundaries of your own organisation, whether building energy management systems, smart grids or e-mobility. It’s not just about the business case, but about creating more value off the back of that,’ Spring remarks.

He admits: ‘The challenges around the energy trilemma are pretty well known. However, the opportunities come from recognising that exponential technologies such as AI, IoT, blockchain and quantum computing are integral to solve the problems around network flexibility and the need for shared value for different participants in the energy system. It’s all about democratising energy with some unusual business models.’