Hydropower's future outlook: improving production forecasting through AI
Following a recent IEA report on the need for substantial investment in both new and existing hydropower generation, Director of Communications, Morten Munkejord takes a look at how hydropower is made more profitable through the use of Artificial Intelligence.
This blogpost is authored by a member of Montel's content marketing team.
The average hydropower plant in Europe is now 45 years old and, generally, in need of modernisation. However, according to the International Energy Agency (IEA), the world's energy watchdog, the "business case for this ageing fleet of hydropower plants has deteriorated due to declining electricity prices and lack of long-term revenue certainty".
Even though hydropower currently accounts for around 30% of the world’s flexible electricity supply capacity, more is still needed to help balance the intermittent output from other renewable sources like solar and wind.
Over the next 10 years, the IEA expects some key developments for hydropower:
Turkey’s hydropower development is expected to drive the largest expansion in capacity in Europe over the coming years.
In North America and Europe, modernisation work on existing plants is forecast to account for almost 90% of all hydropower investment this decade.
Reservoir hydropower plants will account for half of net hydropower additions through 2030.
Run-of-river hydropower will remain the smallest growth segment – mainly because it includes many small-scale projects below 10 MW.
Based on that outlook, it’s clear that in order to help hydropower realise its full potential, higher profit margins for plant operators are key. This is where Montel AI is already playing a role, with a proven track record of lowering error margins (MAE) for production volumes (expected vs delivered) from both run-of-rive and wind power plants, as well as forecasting inflow to reservoirs.
The more precise a producer’s forecasts for near-term production are, the less penalties they will incur in the balancing market. And with multiple weather variables impacting future output, more advanced data-crunching means less errors in these production forecasts.
It's often asked whether AI can be trusted to deliver more precise production forecasts than more traditional models – and the answer is yes, in most cases. And the logic as to why this is the case can be compared to a game of chess:
If machine learning technology can beat the best chess players in the world – not every time, but over time – why can’t it beat the more traditional forecasting models for wind and run-of-river production?"
That said, this is not a game anyone can start playing. That is why Montel AI’s hands-on approach is important. We offer a 3-step process to help anyone improve their production forecasts:
The producer provides historical production data for any given plant and/or unit.
Montel AI inputs the data into the machine learning models.
The models learn over time and produce incrementally more precise production forecasts.
Furthermore, expected production volumes can be reported directly to the spot exchange via standard systems already in place with most operators, lowering the risk of ‘fat finger’ mishaps leading to false reporting.
Power producers from Norway to Turkey are currently using Montel AI to improve their production forecasts, facilitating further green power expansion.
If you’re interested in learning more about how we can help make your wind, run-of-river or hydro reservoir plant more profitable, get in touch!