Sizing a small off-grid power system from real weather
A 2009 paper with Bernard Stark at the University of Bristol: how much solar, wind and battery does an off-grid site need, given its own weather?
Clark, J. D. and Stark, B. H. (2009). “Correlating site-specific meteorological data and power availability for small-scale, multi-source renewable energy systems.” International Conference on Energy, Environment and Sustainable Development (EESD), Paris. World Academy of Science, Engineering and Technology, vol. 54, pp. 754–761. Also published in the International Journal of Humanities and Social Sciences, 3(6), pp. 680–687.
The paper at WASET · Record at the University of Bristol
I built a Matlab and Simulink model of a small, off-grid site powered by solar panels, a wind turbine and batteries. Fed a year of real hourly Met Office weather for a given place, it reports how often the system would fail to meet a steady load, and what the components would cost. Running it across many combinations of panel area, turbine size and battery capacity shows the trade-offs at a glance.
Two findings stood out. The best mix is different for every site, because it depends on that site’s own pattern of sun and wind. And the last 1% of reliability is expensive: in some cases, accepting 99% availability instead of 100% roughly halved the capital cost.
More detail
- The model. Each part — photovoltaic array with maximum power point tracking, wind turbine, battery — is a separate Simulink block, so parts and control methods can be swapped and compared. For each hour the question is simply: was the demand met?
- The weather. Hourly data for 2007 from the Met Office MIDAS archive, treating each weather station as the site. Results were run for eleven UK sites; the paper shows four, from Suffolk to an urban site near the Severn Estuary.
- The costs. Capital costs only, assuming 25-year lives for panels and turbines and battery replacement every five years.
- An example. At one inland site with light winds, a 2.5 kW turbine, 45 m² of panels and 50 kWh of batteries met the load 95% of the time for about £23,000 of components. Halving the downtime was cheaper by doubling the panels than by quadrupling the batteries.
- What it left out. The load was constant. A real load varies by day and season, and a grid connection that can buy and sell would change the answer again.
The title in the Bristol record reads “Correlating historic site-specific…”; the published PDF drops “historic”.