to the grid
Cheaper, Greener, More Reliable
Over the last few years, people have got used to looking at more than one screen at a time.
As you’re reading this article, you might be keeping half an eye on the TV, or occasionally flicking your attention to an email that drops into your inbox. It’s OK – we won’t take offence. You can concentrate on two things at once, because humans are good at multi-tasking.
But imagine having a dozen screens in front of you. All feeding you information constantly, 24 hours a day. A strong wind is blowing in the Irish Sea. There’s low cloud over Portugal. A new industrial plant in Germany is about to switch its machinery on for the first time. A penalty shoot-out in the Italian cup final is approaching its climax.
And you need to act – now. You are an energy trader, and your job is to balance the electricity being generated with the electricity being used. How do you factor in the increased generation from the wind farm in the Irish Sea, the drop-in output from the solar plant on the Algarve, the rising baseload demand in Dusseldorf, and the sudden dip as millions of football fans in Rome switch off their TVs and celebrate in the streets? What do you do?
Dr Aidan O’Sullivan, head of the Energy and Artificial Intelligence Group at University College London’s Energy Institute, has the answer: algorithms.
These lines of machine code are designed to process an extraordinary amount of data, and find the best and most efficient solution to a problem in a matter of milliseconds.
And with the help of those algorithms, you can deliver energy that is cheaper, greener and more reliable.
Energy markets used to be relatively straightforward. Power plants fuelled by coal and gas delivered steady supplies that could be quickly scaled up or down, and the uncertainty lay in predicting the demand. But renewable energy is much less predictable, because so much of it depends on the notoriously fickle weather.
This weather-related intermittency increases what is known as the “imbalance risk” dramatically. To ensure customers get the energy they need when they flick a switch, traders are constantly sourcing spare supply to make up for shortfalls, and offloading any spare capacity that can’t be stored.
Cassim Mangerah, who jointly heads Centrica’s Energy Marketing & Trading business, says artificial intelligence is the key.
“As market trade becomes increasingly exposed to weather, and as we trade very close to actual delivery, we are investing a lot of resources in the development and implementation of algorithms and automated trading systems to optimise balancing and market price exposure."
It is actually only thanks to advances in algorithms that the fluctuating supply generated by many renewables can be successfully integrated into an energy market at all. Otherwise, the balancing act would be too complex for a human to manage.
Algorithms aren’t just helping manage the balance between supply and demand – they are reducing the amount of energy we actually need to generate at all.
Dr Aidan O’Sullivan believes that “the best fuel is efficiency”. The data-processing power of algorithms is spotting efficiencies that once again would be beyond the scope of a human to calculate.
And there is ample scope for further applications.
One example Dr O’Sullivan points to is curtailment, caused when a renewable source generates too much energy in the wrong place, resulting in the need to switch off the asset.
Although advances in storage technology are helping to address this, much of this energy is still wasted, often because the transmission grid is at full capacity and can’t absorb any more energy. And to make matters worse, that wasted energy must be paid for too. It is a problem perfectly suited to artificial intelligence.
“Getting that lost energy back into the system represents a ‘zero investment’ return. We’re already generating that energy, but we’re not able to use it due to poor management of the assets that we already have. So it’s almost like a free lunch.”
The potential benefits are enormous, on both the supply and demand side. Many companies are already using artificial intelligence that can analyse how their machinery operates, and optimise it to use less energy. Some trials have shown efficiencies of as much as 40%.
For wind farms, algorithms can predict when a turbine needs maintenance, and how fast blades will turn at different wind speeds. With ever improving weather forecasting, operators are getting better at predicting energy output as much as 36 hours in advance.
That comes as no surprise to the team at Centrica Energy Marketing and Trading's Aalborg office in Denmark – formerly known as NEAS Energy - which boasts extensive experience working in the liberalised German and Nordic energy markets. Head of Corporate Affairs Rasmus Buhl Møller says algorithms will have a dramatic effect in embedding the advances of the actual physical machinery in our energy system.
“These trading technologies can deliver more value out of renewable assets and are more cost-efficient than building out grid infrastructure. Paired with a virtual power plant approach and investments in flexible generation and storage capacity like the Terhills battery project, they represent a new way to manage energy systems.”
The Certainty Dividend
Renewables need these advances in trading technologies, because the way they are being funded is changing. Historically, preferential treatment helped to insulate the nascent renewables sector from many of the challenges facing new power plants. Now that the technology is proven, it is no longer being given as much assistance.
“They enjoyed generous subsidies and were given priority over other energy sources when feeding power into the grid,” explains Cassim Mangerah. “Times have changed and now renewables increasingly need to play by the same rules as other types of generation. This is where we step in.”
Centrica is playing an increasingly important role in getting new renewables projects to market, by providing what the people who spend money on them, and buy electricity from them, crave most – certainty.
Along with the price of power, winds and sunshine can be volatile, but revenues, power supplies and return on investment need to be reliable.
The energy sector addresses this issue through Power Purchase Agreements (PPAs). A renewable generator promises to supply a certain amount of energy, a large consumer agrees to buy that energy at a certain price, and energy companies like Centrica act as the guarantor in the middle.
They take on the imbalance risk and use the market to smooth out fluctuations in supply.
It’s an approach that opened a route to market for the 650 MW wind installation Markbygden ETT in Sweden. A 19-year PPA with Norsk Hydro, one of the world’s largest aluminium producers, gave the wind farm assured revenue, and Norsk Hydro reliable clean energy.
“As power constitutes around a third of production costs for primary aluminium, securing competitive and long-term PPAs is important,” says Norsk Hydro’s Head of Energy Commercial Tor-Ove Horstad. “But further than price and longevity, we believe green power will give us a competitive advantage in the aluminium industry.”
Other projects are following a similar route. Centrica has signed agreements with a range of generators including solar plants in Denmark, and the backers of a proposal to generate power from a tidal lagoon in Swansea believe they can secure private funding through PPAs after government subsidies were refused.
As Rasmus Buhl Møller points out, a successful PPA is a win for everyone. “PPAs provide stable cash flows and bankability for investors in renewables. And ultimately PPAs will lower the power price for consumers and lower the carbon footprint from energy generation.”
The Human Touch
So does this mean anyone with a smart enough computer can be an energy trader? Not quite, although we may all end up being more than just consumers in the energy market.
Artificial intelligence is only as good as the programme it runs on, and the information it’s given. A poorly designed system can end up producing unintended consequences, which could prove costly not just to individual stakeholders, but the energy grid as a whole.
Rasmus Buhl Møller stresses the importance of human factors like experience and expertise. The algorithms used by Centrica Energy Marketing and Trading in Aalborg are constantly tested, monitored and improved. No detail is overlooked – there is even a weather desk of trained meteorologists operating 24 hours a day, and analysts predicting how every individual wind turbine and solar panel under management will operate under any given conditions.
But the hope is that AI will be able to act as an autonomous trader in local energy communities. As Dr Aidan O’Sullivan explains, matching the supply and demand of your electric car and fridge with your neighbour’s television and solar panel is just the job for AI.
The savings, both in cost and energy, could be unprecedented. So we will be able to keep multi-tasking and using our second screens, safe in the knowledge the machines have it all covered.