A Solar Powered Computer Lab for Malawi

Project Background

Since 2013 The Turing Trust has been delivering computers to several sub-Saharan African countries for ICT skills training in schools and colleges, initially to Ghana and then spreading to Liberia and Malawi. The Trust was established by James Turing, an Edinburgh University student and great-nephew of Alan Turing, the code breaker and computational pioneer. Donated computers are securely disk-wiped, refurbished, and reloaded with software, then transported in shipping containers along with other IT equipment. Together with partner organisations in those countries, teachers are trained and supported in maintaining and using the equipment. Because many remote areas do not have a grid electricity supply, there was a need to provide much more than computers alone: the answer was to design a solar-powered computer laboratory, the “SolarBerry”.

 Basic Design

The original design, by four engineer members of the Rotary Club of Currie Balerno in Edinburgh, was based on reusing a shipping container to house a suite of low consumption Raspberry Pi computers and avoiding the use of inverters that provide an AC supply from DC solar panels. There have been a number of similar solar-powered systems designed not all of which have been successful. Some used conventional mains-powered PCs which required a significant power capacity or had to be specially designed to run on low voltage and without fan cooling. To create mains voltages from relatively low-voltage DC solar supplies requires the use of inverters which are relatively inefficient (typically 80%).  Some IT devices then need adaptors to bring them down to lower operating voltages introducing yet more inefficiency. Our design uses all low voltage equipment obviating the need for inverters and employs low-consumption Raspberry Pi (RPi) computers which require only a 5v DC power supply and use only 4 watts for each device.  The RPi computers are used for both Workstations and a Server, and other devices on the network are also low-voltage designs.

The alteration and fitting-out of the shipping container was effected in Malawi to a design that incorporated plenty of airflow and light by fitting wide double doors to an aperture in one side, retaining the end-doors for access to a service area alone.

The insides were lined with locally-obtained board panels with electrical wiring installed behind the panels or in exposed trunking for neatness and protection.

Interior lighting was provided by 6 x 12v LED ceiling fittings providing good illumination at worktop level yet offering very low power consumption. Benches were created along 3 sides of the unit to provide space for keyboards and mice with monitors being placed at each of the 11 RPi workstations. A compartment for server and battery was constructed across the width of the unit at high level. Cooling of the unit and server cabinet was supported by fitting 4 x 200mm 12v fans at each of the corners (which were initially driven by their own unregulated 50W solar panel until it was found that the high voltages generated in African sunshine could burn-out the fans). To maintain reasonable temperatures inside the unit, shade cloth was draped on steel wires over the roof (except the solar panels) and partly down each side.  Environment engineering calculations showed that this was the most effective configuration rather than tenting right down to the ground. White paint was applied to all external surfaces to reflect heat as much as possible.

Solar Panel and Battery Sizing

The solar panel sizing was estimated from the PVGIS modelling tool, developed by the European Commission’s Joint Research Centre in Ispra, that provides free on-line access to solar radiation and temperature data for PV performance assessment in Europe, Africa and many other areas. The software enables the monthly electrical output to be estimated for any size of PV array at a chosen location including electrical losses, with options for fixed or tracking mounts and for either free-standing or building-integrated. The orientation and tilt angle for a fixed mount may be optimized or pre-selected. Another option enables off-grid performance to be estimated but requires inputted data for battery capacity and its discharge cut-off limit, and the daily electrical energy consumption. (More elaborate modelling software provides shorter interval outputs than the monthly averages produced by PVGIS and may include equipment databases: our results were checked by a commercial user of PVsyst, giving much the same results.)

PVGIS showed that for a 1kW horizontal array at Mzuzu in northern Malawi, the electrical output per month was ~100-160 kWh, the highest output being in October and the lowest in June. This included ~24% losses (for cables, inverter, local high temperature and optical reflectance). (Building-integrated arrays tend to run even hotter, being without a rear air flow.) The annual output was ~1440 kWh, from ~1900 kWh m^-2 of solar radiation. A horizontal array, being less visible, would provide greater security against tampering and theft, but the optimum angle of tilt at this latitude is 13^o which delivers a slightly increased output; a higher tilt than this was used in the installation to enhance the winter performance.

The starting point for determining PV array and battery sizes is an estimate of the electrical energy required to operate the demand load, which requires power ratings and operating periods for each item of equipment. The electrical items are tabulated below, along with their power rating and current, and all items use 12V DC. In addition there is an initial estimate of the “worst case” daily hours of usage. Hence the total daily consumption is almost 5kWh. This software does not distinguish between weekdays and weekends although two separate estimations could be run for this. Ignoring any difference in daily usage, the annual consumption would be ~1800kWh.

Another iteration is required to enable the battery size to match the load demand. Using the off-grid option in PVGIS requires inputs of PV array size (initially set at 1kW, but in practice 6 x 150W panels were installed), battery capacity (initially set at 320 Ah, but in practice 4 x 100Ah were installed), discharge cut-off limit (default value is 40% but can be lower depending on battery type and construction), and daily electrical consumption. The PVGIS model estimates the electrical energy generated per day and the battery status at the end of each day, both given as an average per month. Just as for the estimate of PV array size, the battery modelling shows a deficit in the initial chosen size: this could be addressed by increasing the sizes, but a more realistic estimate of the load usage shows that these PV and battery sizes would provide the electrical power demand, thus avoiding increased costs.

^** More details in the SolarBerry Information Pack.

^{^^} Use in evenings for films.

^##For occasional use only by items that demand a 240V AC supply.

Additional Equipment

The main battery charge controller, to prevent overcharging and reverse current flow, was a Solar 80 PWM 80A capacity unit. This performs almost as well as a more expensive MPPT type in hot environments: PV arrays give a lower output voltage at maximum power point for temperatures above a nominal 25^oC and the benefit of a voltage-adjusting MPPT controller is lost. Optimum strategy for the controller (i.e. priority with regard to charging the batteries) relies on a battery monitor system to give full information on charge state of the batteries. This has been designed, built and tested but has still to be implemented due to resource constraints. It uses another Raspberry Pi computer and an Arduino A-D converter to monitor and record voltage and current from the PV array, battery charge state, solar irradiance and ambient temperature. The data are available in graphical format in a browser page.

Noting that PV arrays cannot be switched off and are always live, especially in daylight hours, electrical isolation was provided before and after the battery charge controller and at each computer station, as well as for the lighting circuit and other components by double-poled DC isolators. Miniature DC circuit breakers were also installed in the main consumer unit.

To cater for Movie Nights DVDs can be played and projected either indoor or outdoor using a low-energy LED projector with sound played through a 100W PA system.  The provision of entertainment as well as phone and tablet charging allows a small income to ensure that future maintenance costs can be covered.

Project Outcome

The prototype SolarBerry has proved the viability of the concept and has been successfully installed in a remote community at Choma near Mzuzu in N Malawi which had no mains power and no prospect of being connected in the foreseeable future. The unit has been enthusiastically received by the local community which comprises two Secondary schools and a Primary School and since its installation in July 2018 has provided both child and adult education.  The solar design has proved more than capable of sustaining the required power demand of the unit. The project has also proved invaluable in gaining experience in solar generation in rural African communities for rolling out future projects.

Resources

1. Photovoltaic Geographical Information System: https://ec.europa.eu/jrc/en/pvgis

2. The Turing Trust: https://turingtrust.co.uk/home/our_work/solarberry/

3. Information pack for non-commercial use: https://turingtrust.co.uk/home/our_work/solarberry/how/

Acknowledgements

Jim Douglas and Ian Campbell contributed their expertise to the design along with the co-authors, and Sam Gray designed the battery monitor circuitry. Brian Ferguson managed the building of the Malawi unit. The Institute of Physics is thanked for the award of a Virdee Grant. Our thanks also go to the Scottish Government’s International Development Small Grant that provided all important funding for this project.