The journey

Alexis arrived when the equipment had been ordered but nothing was connected yet. His job was to turn a box of parts into a working installation. He fixed the SMA Sunny Boy inverters vertically on painted plywood boards (to keep the clearance they need to cool down), drilled the cable holes through the floor rather than the walls to protect the airtightness of the passive house, and built a small shed to protect the firefighter switches from the Bergen rain.

His main creation is the patching board, an idea worked out with Harald and Richard and inspired by old telephone switchboards. It lets students redirect each string of panels to an inverter or to the battery with plug-in cables. Alexis drew the plates in Autodesk Inventor for the plasma cutter, added colour-coded labels for teaching, and calculated which combinations of panels stay within the voltage and current limits of each device.

He wired the DC side and left a clear plan for the next steps: the AC side with an electrician, the Tigo monitoring devices, the flexible panels, a screen visible from the window, and a first sketch of a new logo.

Gaétane's semester was shaped by the COVID-19 pandemic: Norway closed its borders, and she carried out the whole project remotely from France. She still left three useful pieces of work.

She finalised the SolarLab logo in Adobe Illustrator, in two weights (one readable at small size) and with a design that works in both English ("Solar Lab") and Norwegian ("Sol Lab"). She then designed a first live display: a Raspberry Pi reading the Sunny Boy inverters over Modbus TCP, a screen showing daily and yearly production, and a database to keep several years of data.

Finally, she studied the SMA Sunny Island, the battery inverter that would let the house run as its own small grid, and chose a single-phase setup that respects the 3 kW limit and keeps the three existing inverters.

Julie connected the Tigo optimisers, the gateway and the cloud logger, and got the first per-panel production data. When the logger failed, she worked with Tigo support until it was replaced under warranty. She also moved the lab to one-minute data instead of fifteen minutes, which made a clear difference on Bergen's changing days.

The first data raised the questions that later students would work on: production much lower than expected in November and December, a battery unit that emptied in one or two days, and a weak Wi-Fi signal through the insulated walls. Julie built an Excel database of University of Bergen radiation data from 2010 to 2021 to compare theory and reality, and showed that October 2021 was a record rainy month.

She then designed the AC side around the Sunny Island (battery, generator and load management), prepared the fuse box with an electrician, and studied an EV charger. The components were ordered just before the end of her semester.

Alexandra started with a clear list of open problems: the battery that emptied too fast, the weak internet connection, and production far below the theory. She found that the two strings of panels did not produce the same amount (sometimes the right string made only 42 % of the left one) and suggested shading or Tigo settings as possible causes.

She compared real production with two simulation tools, PV*SOL and the European PVGIS, and prepared the installation of the Sunny Island (space, weight, mounting). She also researched how to measure solar radiation and proposed a Kipp & Zonen SMP11 pyranometer, the same family as the University of Bergen's reference instrument.

Alexandra also worked on how the lab presents itself: information panels for visitors, a screen behind the window, an updated logo with the university name, and a study of bidirectional car charging (vehicle-to-home).

Pierre's first task was to find out why the Sunny Island would not start. With Harald, he tested the communication cable and found the real cause: the PowerOak battery was not on SMA's list of compatible batteries, so the two devices could not talk to each other. He then ran a market study and a sizing calculation for a new lithium battery.

Most of his semester was spent on simulation. He modelled the campus, the neighbouring buildings and the Ulriken mountain in SketchUp, and used the Skelion plugin with PVGIS to calculate shading losses. When moving the lab to another part of the campus was discussed, his results showed that the other site would be shaded around ten times more, and the move was put on hold.

He also chose the place for the pyranometer: on a pole on the roof of the main building, the least shaded spot on campus, with the pole sized against Bergen's strongest recorded wind.

Tom taught himself Python to make the pyranometer talk. He set up a Raspberry Pi, communicated with the sensor over Modbus RTU, and wrote a program that saves the measurements to files while plotting them live, which took several attempts to get both running at the same time.

He then compared the lab's irradiance with the University of Bergen's station (Florida): the curves follow the same trends, and his work showed how much the sampling interval changes what you see. He also found and corrected a two-hour time offset in the reference data.

In the second part of his project, Tom calculated the heat demand of the passive house and the battery capacity it needs, and compared three SMA-compatible batteries (BYD, LG Chem, Tesvolt), preparing the purchase made the following semester.

Alexis started from Tom's list of suppliers, contacted seven distributors and compared price, reliability and distance. The choice was a Tesvolt TS25 battery from the French distributor Perma Batteries; it was ordered during his semester but had not arrived when he left.

He improved the pyranometer program so it draws three graphs (one week, three days, one day) and compares the measured irradiance with theoretical clear-sky curves. After trying to host a site directly on the Raspberry Pi, he bought the domain and built the first solarlab.online with Joomla, with the Raspberry Pi sending the graphs to the server automatically.

He also fixed many day-to-day problems, like connecting the Raspberry Pi to eduroam with an external antenna, and wrote a section of his report specifically for the next student.

Lucie focused on making the lab usable by others. She wrote a system management notice explaining how to monitor and configure every device, and step-by-step instructions to restart the battery and the Sunny Island after a shutdown.

She worked on the data too: she calculated the best panel tilt for each month of the year and the losses at 30°, 45° and 60°, compared the lab's irradiance with the University of Bergen's, logged car charging sessions, and started a monthly table of PV production, generator use and consumption.

She also updated the electrical diagram of the whole installation.

With the days getting shorter, Esther's main challenge was the backup generator (in practice the campus grid). She followed the system daily on Sunny Portal, worked on the relays that let the Sunny Island start and stop the generator, and wrote a report for SMA customer service about the faults.

She programmed the Flexit heat recovery ventilation to run harder during the day and slower at night, to save energy without losing air quality. She also started a full-year view of the pyranometer data, and simplified access to the devices by changing their IP addresses.

Esther also finished the interior: she fitted the wooden ceiling panels around the ventilation pipes, with the help of Nafez from the research team.

Theo found two blind spots. First, nobody could tell how much energy came from the generator, because the portal showed one mixed total. He added a dedicated energy meter so the generator's share is now measured. Second, when the battery reached 95 %, SMA stopped the production and data was lost. He installed a Wi-Fi relay (Shelly) on a radiator, so someone could switch it on remotely to use the surplus.

He then gathered a year of data from Tigo, the pyranometer and the SMA portal, cleaned it with Excel macros, and compared it with PVGIS simulations. The comparison gave monthly correction factors for better estimates.

His report also includes a clear summary of the lab's history in six phases, and recommendations for a smarter, automatic energy management.

Simon took Theo's manual relay and made it automatic. His Python program reads the battery state of charge from the Sunny Island over Modbus TCP and switches the heater through the Shelly relay, with a manual mode as a fallback.

He then extended the idea to all the controllable loads. Instead of fixed thresholds, he introduced energy levels (EL0 to EL3) with hysteresis, so loads do not switch on and off too often, and the car chargers come before the heater. He designed and built the relay box for the two car chargers and the heater, and 3D printed supports and socket covers.

He also fixed the website graphs, which had stopped updating: the uploads used up the monthly data allowance, so he reduced them to once an hour and made the program reload the last seven days after a restart.

Bastien noticed that production flattened at about 1,000 W on sunny days instead of rising to 1,600 W per array: the inverters were saturated. By rewiring the panel strings, he raised the maximum output from 2,000 W to 3,000 W. He also built a production forecast model from the sun's position, the panel orientation and the radiation, first linear and then polynomial.

He put the automatic energy management into service, with a small interface, and tackled the lab's crashes: a backup power supply keeps the pyranometer and the website running during a blackout, and an auto reboot script restarts the system on its own in about ten minutes. For the first time, the SolarLab could manage its loads and recover from a crash by itself.

Bastien also created the solarlab@hvl.no accounts, made backups of the Raspberry Pi, presented the lab's data to another HVL class, and listed ideas for the future, like a partnership with the HVL hydrogen lab.

The old solarlab.online was rebuilt from scratch so that the lab's data is live, open and downloadable for students and researchers.