Fion's battery storage system is equipped with AI. What exactly does the AI do?
Johannes Meriläinen: It continuously calculates which application and which mode of storage deployment is most economical and sensible for each customer installation. Often two applications are theoretically viable from an economic standpoint, but one requires less storage capacity and therefore delivers a higher return per cycle. The AI thinks, analyses and optimises continuously, determining how the storage system should be operated in every 15-minute interval, which is the standard settlement unit in the energy market.
What data does the AI need to control the storage system in line with local requirements?
The AI needs several data points. On the one hand, we read out a wide variety of meters installed at the industrial customer's site, covering generation assets, individual loads and transformer stations. On the other, market prices and weather data are analysed. In addition, we forecast consumption and generation in order to determine the storage schedule and the capacity available at any given time.
Stay informed on the solar energy transition – subscribe to our free newsletter
Fion Energy
Does it also draw on weather forecasts and similar data for the solar system?
Yes. Weather data feed into the control system, in particular to forecast the output of the solar installation and align storage operation with it.
Where does the system get the data from?
The storage system has its own control box on site, which communicates with our software in the cloud. The computationally intensive optimisations run on our high-performance servers in Germany. The storage system receives the calculations and control commands from the cloud. If the connection is interrupted, the storage system still knows what to do: it then prioritises defined applications such as peak shaving or self-consumption optimisation, depending on the customer case.
Can the company's production schedules be entered manually, for example, so that the storage system adjusts its operation accordingly?
That is not necessary. We design the storage system so that it works on the basis of historical load profiles. These typically already cover different weeks with different production schedules. The key point is that industrial companies do not need to adapt their production. The storage system is an added layer of flexibility that sustainably reduces peak loads and shifts loads in a grid-friendly way, saving on grid fees.
Greece – Renewable & Storage Forum dates updated to 21-22 October
How quickly does the AI react to changes in the load and generation situation?
The control system reacts every second, like a heartbeat, and in some cases even in the sub-second range. This means the storage system's operation is continuously adjusted to the current situation.
How exactly does this adjustment work?
Through situational prioritisation of applications. If the storage system is responsible for both peak shaving and self-consumption optimisation, peak shaving takes precedence. A single peak over the course of a year can permanently increase grid fees. Control is always based on the measurement at the grid connection point: the system continuously reads what is being drawn or fed in at the connection point and operates the storage system so that a defined setpoint is not exceeded in either direction.
Harry Wolkenfelt: Insurability is built, not bought
What additional business models does AI-based control make possible?
The marketing of flexibility, for example. At times when the storage system is not needed on site, we pool it with other Fion storage systems to form a virtual large-scale battery and market this on various exchanges, such as spot intraday or day-ahead trading, to generate additional revenue for the customer. We see further models emerging in the future. If grid fees become dynamic, for instance, the storage system can factor in those dynamic signals as well and reduce grid fees even further.
Does AI implementation make it possible to serve several business models at the same time?
Yes, that is precisely what our AI control system does. It combines several applications such as peak shaving, shifting the use of solar power and taking advantage of flexible electricity tariffs, and prioritises them situationally to deliver the best overall economic result.
Spain – UNEF calls for more flexible rules on data centres
How is the storage system sized when several applications are used simultaneously? Peak shaving has different power and capacity requirements than shifting solar power use.
We size the storage system on a data-driven basis and deliberately avoid oversizing it. Based on the customer's current and historical energy data, we simulate different storage sizes and identify the one that delivers optimal savings alongside an optimal payback period. As a rule, peak shaving is the decisive factor for sizing, not self-consumption optimisation. This means we define exactly the right storage size and the intended applications for each site, rather than oversizing the system.
What savings are possible with such a system?
Savings are highly site-dependent. To determine them for each site, we offer an independent, data-driven and non-binding analysis, including the appropriate storage size and the intended applications. Payback periods of three to five years are typical in the market, with outliers in both directions. That makes the investment very attractive. We have customers whose storage systems deliver annual savings in the six-figure range, depending on the size of the system.
Interview by Sven Ullrich.
