Uncertainty & Externalities¶

Scenarios 6–7 · demand grows +5 %/year · on-grid · two futures

The previous scenarios assumed the future was known. These two relax that assumption. First we let the timing and reliability of a future grid connection be uncertain, using the stochastic formulation of MicroGridsPy. Then we internalize an environmental externality by pricing carbon emissions inside the objective.

Scenario 6 — Stochastic grid connection¶

The demand trajectory is the same 5 %/year growth as before, but the future context of the mini-grid becomes uncertain. The model considers two equally probable futures:

• Scenario 1 (on_grid_5) — the national grid arrives in 2030;
• Scenario 2 (on_grid_7) — the grid arrives later, in 2032.

Crucially, the grid is not assumed to be perfectly reliable. Availability is represented hour by hour, with outages that gradually diminish as service quality improves. The earlier connection (on_grid_5) starts less reliable (more, longer outages) than the later one.

Fraction of hours the grid is available, by year. Before the connection year availability is zero; afterwards it is high but imperfect, reflecting the modelled outage process (see the grid-availability methodology).

This is a two-stage stochastic problem with recourse: MicroGridsPy chooses one system design (shared across both futures) but lets operation adapt to each future separately. The dispatch makes the difference tangible — in 2030, the early-connection future can already import grid electricity in the evening, while the late-connection future still runs as an isolated mini-grid and covers the evening peak with diesel and storage.

Average-day dispatch in 2030. Left: grid has arrived — evening load is partly met by grid import (blue). Right: still isolated — the evening peak is met by battery discharge and diesel.

Because the system no longer has to rely entirely on local generation and storage for the whole horizon, the expected economics improve relative to the off-grid capacity-expansion case:

Operating cost for each of the two futures, and the dashed expected value weighted by their probabilities. Each future has a different cost path — stochastic planning finds a single design that performs well across both.

Scenario 7 — Carbon cost¶

The final scenario keeps the same two-future grid structure but adds an environmental signal: carbon costs are included directly in the objective function. Emissions are priced at 0.1 USD/kgCO₂e and cover three scopes:

• Scope 1 — direct emissions from local diesel generation;
• Scope 2 — indirect emissions from imported grid electricity (grid emission factor 0.35 kgCO₂e/kWh);
• Scope 3 — embodied emissions from manufacturing and installing the components.

With emissions priced, the model shifts toward a cleaner configuration: it installs more solar PV and runs the diesel generator less.

Pricing carbon raises expected NPC and LCOE by about 6 %, but the model responds by building a more renewable system that reduces the emissions it has to pay for. The emissions themselves are now an explicit output, resolved by scope and over time.

Emissions by scope and year. Scope 3 (embodied) is concentrated at the start, when components are installed; Scope 1 (diesel) dominates the early off-grid years; Scope 2 (grid import) appears and grows once the grid connects.

Where to go next¶

These six scenarios show how progressively richer assumptions — technology, growth, staging, uncertainty, externalities — change the least-cost design of the same mini-grid. To reproduce or extend them, start from the reproducibility mapping, and see the Methodology for the equations behind each feature and the User Guide for the practical workflow.