CapFor Online for Fingrid: taking power generation capacity forecasts to a whole new level

Finland is one of the forerunners of Combined Heat and Power (CHP), with almost one third of its electricity generation coming from CHP plants. The market share of electricity produced from CHP is unmatched by any other country. Nuclear power is another essential element of Finland’s energy system, accounting for a 33.7% share of domestically produced electricity.

The electricity generation in the country is quite distributed among different sources, with an increasing amount of intermittent generation from renewable sources being connected to the grid in the last years. Balancing the needs for clean, reliable and efficient energy sources within Finland, and coordinating its operations with the neighbouring Nordic countries is a though challenge for Fingrid, one that requires expert solutions.

The Nordic Regional Security Coordinator (N-RSC) is an organ created to help coordinate the operations of the TSOs in the Nordic countries, and started its operations in 2018. It ensures the system is well in harmony and effective, providing a common ground for information exchange between its members. To do so, N-RSC requests that every one of its members, Fingrid included, present a 7-day forecast of the total generating capacity of their power plants.

Therefore, Fingrid needs to have a clear idea on how much energy is available to be produced within the country. However, finding this information is not so straightforward – the energy that is effectively produced is seldom the same as the available capacity. As such, we only have records of the actual outputs of the power plants, not their potential, and with this comes a huge challenge for the grid operators. Having a clear idea about the available capacity is essential in running the Finnish grid efficiently and reliably.

This challenge was the main driver behind the development of CapFor Online, a software platform that forecasts the available capacity of all CHP and nuclear power plants in Finland for a 7-day horizon. It was posed as a big challenge, given how the electricity production from these sources is dependent on many factors. Nuclear plants have scheduled maintenance cycles and a more predictable operation, but CHP plants have their available capacity fluctuating depending on the weather or electricity price, for example.

 CapFor Online forecasts these dependencies, enabling Fingrid to assess the availability of two of the main energy sources in the country. This allows them to be prepared for possible adequacy solutions beforehand, and properly plan for the energy demand in the following week. With the increased penetration of renewable energies in the grid, the forecast of non-intermittent sources such as nuclear and CHP becomes even more important for ensuring the grid stability and adequacy. Being able to accurately forecast how much electricity Finland is able to generate is critical to keep the lights on and the country running efficiently.

The project was born with a short pilot, during which a technology demonstrator was created in close collaboration with Fingrid. After this proof of concept was tested, work began on a more robust version of the software. In just a few months, a completed working solution of CapFor Online was delivered, providing fully online operations and generating output forecasts on a daily basis.

Machine Learning algorithms were used to analyse huge amounts of data and construct unique models for the largest power plants in Finland. Smaller plants, which contribute less to the overall generation figure, were aggregated based on their locations and characteristics. Weather forecast data from the Finnish Meteorology Institute has been integrated into CapFor Online, further increasing the accuracy of the capacity forecasts. Additionally, information from the ENTSO-e Transparency Platform is used to enrich the data with knowledge of the availability of each plant. This allows for the creation of extremely accurate predictions of their capacity and contribution to the grid.

When operating the electrical system the forecasts have become more and more important. The forecasts made by Capfor Online help us to get better overlook of the adequacy situation for the upcoming days.

Paavo Pietilä, Specialist, Fingrid Oyj

However, not all power plants are created – or forecasted – equally. Some have such stable outputs that their available capacity is easy to predict, while others present quite flexible behaviours, with large variations over time. Building accurate Machine Learning models for these flexible plants was one of the biggest challenges in creating CapFor Online. VTT combined its expertise in Artificial Intelligence and in Power Systems to come up with a novel and efficient solution: these plants’ available capacity would be modelled using probabilistic forecasting techniques. It was a perfect example of how the combination of different knowledge areas can be used to develop new technologies and solve demanding, world-class challenges.

CapFor Online is currently in use by Fingrid, providing accurate forecasts of energy generation capacity for Finnish nuclear and CHP power plants – solving a challenge that is well known by many grid operators. It forecasts the available capacity of a country’s electricity production in a very precise way – boasting impressively low error rates of just below 5% (depending on plant model and characteristics) – and can deliver daily forecasts up to 7 days in advance.

CapFor Online now supports Fingrid’s operations by improving its communications with N-RSC through enhancing data quality, availability and transparency. With a more accurate forecast of the available generation capacity of baseload technologies such as nuclear and CHP, Finnish electricity grid operates more securely and effectively. Furthermore, the tool’s highly customizable design allows for its adoption in other countries and energy markets.