The Kalobeyei Case Study¶
All the examples in this section use the case of Kalobeyei, a remote settlement in northern Kenya that is currently not connected to the national grid. Like many rural communities in Sub-Saharan Africa, households and businesses still rely largely on traditional energy sources such as kerosene, diesel, and biomass.
This makes Kalobeyei a realistic testbed for MicroGridsPy: it combines a genuine electricity demand, good solar potential, and the kind of planning uncertainty — growing demand, possible future grid arrival — that off-grid planners face in practice.
The community¶
The demand is built bottom-up from the settlement's users:
- around 700 households, classified into three groups by their actual and expected appliance ownership;
- 2 primary schools (roughly 30 lights each, plus some external lighting);
- 1 dispensary with three rooms and seven beds;
- productive activities: tailoring workshops, bars, and garages.
Demand assessment¶
Before running any optimization, MicroGridsPy characterizes the local demand through an hourly time series for the representative year. The daily average profile is typical of a small rural settlement: low overnight consumption, a daytime rise driven by productive uses, and a pronounced evening peak when lighting and social activity coincide.
Average daily load. The evening peak reaches roughly 80 kW, with an average daily demand of about 1100 kWh. In some scenarios (from Planning for Growth onward) this demand is scaled up year by year to represent settlement growth.
Resource assessment¶
The other primary input is the availability of the renewable resource, again provided as an hourly time series. Kalobeyei has strong and stable solar potential, which makes solar PV the natural backbone of the system.
Average daily solar capacity factor. Wind was also assessed for the site but is more variable and generally lower, so it contributes little to the least-cost design; the examples therefore model solar PV as the single renewable technology. The resource time series were generated for the site's coordinates using open resource databases.
Ambient temperature¶
A third time series matters once battery ageing is modelled physically rather than assumed. Both mechanisms in the semi-empirical degradation model — calendar fade and cycle fade — are temperature-dependent, so Scenario 3 needs an hourly ambient temperature profile for the site.
Hourly 2 m air temperature for the site (left) and the average daily profile with its daily min–max band (right). Source: Renewables.ninja Point API, MERRA-2 reanalysis at 3.770 °N / 34.625 °E, calendar year 2019.
Turkana is hot and the daily swing is large: the annual mean is 26.5 °C, but hours range from 18.5 °C to 38.3 °C, with roughly an 11 °C spread between night and mid-afternoon. That variation is not cosmetic — it makes the cycle-fade coefficient vary by more than a factor of two over the year, so a kWh cycled through the battery on a hot afternoon costs materially more ageing than the same kWh at night. The series is one representative year, repeated across the 10-year horizon like the demand and resource inputs.
These inputs — demand, resource availability, and (where degradation is modelled) ambient temperature — define the operating conditions under which every scenario is optimized.
Shared assumptions¶
Unless a scenario explicitly changes them, all six examples use the same techno-economic and project settings. This is what makes their results directly comparable.
| Setting | Value |
|---|---|
| Planning horizon | 10 years (2026–2035) |
| Social discount rate | 7.4 % |
| Formulation | multi-year (dynamic) |
| Max Lost-load constraint | enforced 0 % (unmet demand not allowed) |
| Minimum renewable penetration | not enforced |
| Integer (discrete) capacity units | disabled |
| Land constraint | not enforced |
| Generator efficiency model | constant efficiency (no part-load curve) |
| Battery loss model | constant efficiency |
| Battery degradation | exogenous capacity fade only |
| Diesel fuel cost | 1.3 USD/litre (constant across years) |
Techno-economic inputs¶
| Parameter | Solar PV | Diesel generator | Li-ion battery | Lead-acid battery |
|---|---|---|---|---|
| Capital cost | 1200 USD/kW | 400 USD/kW | 300 USD/kWh | 200 USD/kWh |
| WACC | 5 % | 5 % | 5 % | 5 % |
| Lifetime | 25 y | 20 y† | 10 y | 8 y |
| Fixed O&M (of CAPEX/yr) | 1.8 % | 4.5 % | 2 % | 2 % |
| Efficiency | inverter 0.98 | full-load 0.34 | charge/discharge 0.93 | charge/discharge 0.86 |
| Operating limits | DC/AC ratio 1.2 | — | DoD 0.8; C-rate 0.25 | DoD 0.5; C-rate 0.167 |
| Exogenous degradation | 0 %/y | 0 %/y | 0.5 %/y | 0.3 %/y |
Diesel LHV = 10.14 kWh per unit of fuel. Battery inverter cost = 160 USD/kW; PV
inverter cost = 180 USD/kWac; inverter lifetime = 15 y; inverter fixed O&M = 1 %
of CAPEX per year. †The generator lifetime is 20 y in every scenario except
capacity expansion, which uses 10 y
so that each investment step carries its own generator annuity. Emissions are zero in
scenarios 1–6; the carbon-cost scenario prices all
three scopes — diesel combustion at 2.69 kgCO₂e per unit of fuel (scope 1), grid imports at
0.35 kgCO₂e/kWh (scope 2), and embodied emissions of 500 kgCO₂e/kW for PV and
100 kgCO₂e/kWh for the battery (scope 3). Full inputs are in each project's inputs/
folder — see the reproducibility mapping.
With the site characterized, we can start comparing planning strategies, beginning with Technology Choice.