With the growth of intermittent renewables, machine trading of energy is becoming both more sophisticated and widely practiced.
With the growth in decentralised renewable energies, the energy trading market is becoming a lot more complicated due to their variability and the challenges in forecasting output and demand and the consequent prices.
As an example, Richard Plum, Consultant and Energy Economist at consulting company ProCom, in an interview with Engerati cites the case in Germany, which is occurring with increasing regularity, of demand running at around 70GW and a high proportion of supply, sometimes over 80%, coming from renewables.
“These are big numbers involving a large number of generation assets,” he says. “Consider that for every asset it is necessary to have a day ahead forecast and to update the intra-day forecast every 15 minutes.”
Short of engaging an army of traders the only way such forecasting can be viably done is by machine and consequently the use of machine trading of energy is growing.
The question is: how effective is machine trading and in particular, can it match the ability of experienced traders with deep knowledge of the market with which they can react on a minute by minute basis?
Plum points to two different variants of machine trading – auto-trading and algo-trading – stating that from his experience there is often some misunderstanding of the difference between them.
“Ultimately it comes down to the level of intelligence that is involved,” he says. “Auto-trading has little or no underlying intelligence whereas algo-trading has a higher level of usually machine learning based intelligence.”
Nevertheless, both have their advantages and have their place in the energy trading market, he adds.
For example, auto-trading is generally limited to buying and selling taking little account of imbalance issues or prices or other market trends and is generally utilised by smaller traders who don’t wish to trade around the clock or at weekends.
On the other hand, algo-trading is generally practiced by the larger traders which have the capacity to set it up, although they might also undertake auto-trading as an addition, for example for a speculative trade involving little effort.
“If I am focused on the French market but want to make a dispatch or close a position on the German market, I would give it to an auto-trader so I can concentrate on my main market where more money is to be made.”
What makes algo-trading more sophisticated and gives it its power is the underlying algorithms, which in turn require the expertise of experienced traders to develop.
“Behind every good algo-trader is a trader who fully understands the market,” says Plum.
He adds that as such, algo-trading needs a lot of work to bring to a working status and also needs to be constantly modified as the market rules and regulations change.
“It involves a lot of testing and back testing of algorithms along with the machine learning analysis to maintain to a working status.”
So what is needed to establish an algo-trading platform?
The most important issue is the need for a “very effective IT infrastructure” in the company, Plum says.
“It’s all about automisation. A lot of different data is needed such as weather forecasting and market pricing, and these need to be automated so that the day ahead and especially intra-day forecasting can be done to schedule.
“And if data is missing such as the latest weather forecast, automated processes need to be in place to replace it.”
The second point he mentions is redundancy of systems and trading desks.
“If there is a breakdown, and they do happen, one needs to be able to switch to another system as fast as possible.“
He adds: “The more automation and the more data one can bring in, the more effective the trading will be. And with the algo-trader running in the background, it frees up the human traders to spend more time on the analysis and design of their algorithms.”
With the ongoing growth in variable renewable energies and the further development of the European market, with new interconnections leading to more cross-border trading and strategies such as arbitrage and speculative trade, there is little doubt that machine trading is here to stay.
New technologies such as blockchain also are being introduced in the space, which could bring further automisation and elimination of paperwork to the trading process, although it is likely to be up to five years before it could become mainstream.
So could machine trading supersede human trading?
While the REMIT (wholesale energy market integrity and transparency) rules set out regulations on trading, the extent of future machine trading in Europe will depend on how these regulations evolve over time and the interests of the energy exchanges, Plum says.
He adds that experience from the financial industry shows that there can be “breakdowns” with resultant financial damage from a too great reliance on machine trading in the market.
“While we are moving more and more towards auto-trading and algo-trading, ultimately there always has to be a person behind machine trading as machines can’t do everything on their own.”
And so - at this time at least - is a good machine energy trader a match for its human counterpart?
“Those with the best algo-traders make the money whereas the others lose it,” says Plum.
“From talking to many traders, the general opinion is that algo-traders can do a lot of volume and deals but really good energy traders will tend to beat the algo-traders with their market experience. I’ve heard of just one case of an algo-trader bettering an actual pro trader looking at the profit that was made.
“Ultimately it comes down to how good the traders are and how good their algorithms are.”