The 675 million people the transmission grid will never reach

Roughly 675 million people lacked access to electricity in 2023, according to the IEA's World Energy Outlook, and about 80% of them live in rural areas of sub-Saharan Africa and South Asia where extending the medium-voltage network costs $5,000–$20,000 per household connection. That number is the single most important input to any microgrid business case. When the cost of a grid extension crosses roughly $1,500 per connection, an isolated solar-plus-storage system wins on a levelized basis — a threshold that the World Bank's ESMAP program has used since 2018 to screen mini-grid sites.
The economics have moved decisively in the last five years. Utility-scale lithium iron phosphate (LFP) cell prices fell from about $1,100/kWh in 2010 to $115/kWh in 2024, and small-scale DC-coupled microgrid inverters have followed. A 100 kWp village microgrid that would have cost $600,000 in 2015 now prices out at $280,000–$350,000 depending on storage autonomy, according to IRENA's renewable cost database. Diesel, meanwhile, has gone the other way: delivered fuel prices in landlocked African and Pacific markets routinely run $1.20–$2.00 per litre after transport, which is why diesel mini-grid operators report LCOEs above $0.70/kWh.

What a solar microgrid actually contains

A remote-community microgrid is not a scaled-down utility substation. It is a set of four subsystems that must be sized together: generation, storage, conversion and distribution. Generation is almost always monocrystalline PV — 550–600 W bifacial modules at 21–22% efficiency are the current default for ground-mount arrays, with TOPCon modules reaching 23% in some tenders. The array is deliberately oversized relative to average load by 1.4–1.8× to account for monsoon seasons, dust soiling and high ambient temperatures, which knock 8–12% off nameplate output in the Sahel and 5–9% in coastal Southeast Asia.
Storage is the cost driver and the reliability driver. A typical design carries 1.5–2.5 days of autonomy using LFP racks, cycled to 80% depth of discharge, with a 10-year or 6,000-cycle warranty. Conversion happens at two points: a DC/DC MPPT stage on the array, and a bidirectional inverter that forms the grid voltage and frequency for the community. Modern inverters for microgrid duty run grid-forming firmware, which means they set voltage and frequency themselves rather than following an external reference — a hard requirement for black-start capability and for absorbing the surge of a 5 kW rice mill or a welding set.
Distribution is the part that gets underestimated. Low-voltage three-phase distribution through a village of 200–400 households, with service drops and smart meters, typically consumes 18–25% of total project capex. Prepaid smart meters with remote load-limiting have cut non-technical losses from 30%+ on early mini-grids to under 8% on systems commissioned since 2021.

Diesel, hybrid or pure solar-plus-storage: the numbers

Sizing decisions hinge on the load profile, not on the technology preference. The table below reflects published ESMAP and IRENA cost benchmarks for 50–500 kW village systems as of 2024.
| Configuration | Capex per kW installed | LCOE ($/kWh) | Fuel dependence | Typical O&M | Design life |
|---|---|---|---|---|---|
| Diesel-only genset | $1,200–$2,500 | $0.55–$0.90 | 100% | High, fuel + rebuilds | 5–8 years (engine) |
| Solar + battery + diesel hybrid | $1,800–$3,000 | $0.30–$0.55 | 20–40% | Medium | 12 years genset, 25 PV |
| Solar + LFP storage only | $2,000–$3,500 | $0.20–$0.40 | 0% | Low | 25 PV, 10–12 storage |
The pure solar-plus-storage column is cheapest on LCOE but carries the highest upfront ticket. That gap is exactly what concessional finance exists to bridge. Blended finance structures under the World Bank's $465 million Mini Grids for Half a Billion People program offer 30–50% capital subsidy on the condition that tariffs stay below a regulator-set ceiling, typically $0.20–$0.30/kWh for residential tiers.
Hybrid designs still dominate above 300 kW average load, where a 60–100 kVA diesel unit serving as backup is cheaper than adding another 6–8 hours of battery. Below 150 kW, the trend is strongly toward diesel-free. Husk Power Systems, which operates more than 200 community microgrids across India and Nigeria, has publicly committed to 100% renewable systems and reports 99% uptime with no diesel on site at its newer sites.

Deployment data from India, Nigeria, Indonesia and the Pacific

India's mini-grid sector is the largest in the world by capacity. The country's Saubhagya scheme connected roughly 28 million households between 2017 and 2022, but the last-mile population — around 5 million people in forest and hill districts — remains off-grid by geography rather than policy. Roughly 1,300 operational mini-grids with a combined 90 MW of capacity served those areas at the end of 2024, according to the Mini-Grids Partnership's market tracker.
Nigeria carries the biggest single national gap: about 85 million people without reliable electricity, out of a population near 230 million. The Rural Electrification Agency's Nigeria Electrification Project has disbursed performance-based grants for over 150 mini-grids totaling around 20 MW, with tariffs launched at ₦150–₦250/kWh ($0.10–$0.17). Remote sites in the Niger Delta typically pair a 40–120 kWp array with 100–300 kWh of LFP, enough for lighting, phone charging, irrigation pumps and small agro-processing.
Indonesia and the Philippines present the opposite profile: high fuel costs, thousands of small diesel-powered islands, and monsoon patterns that demand 2.5+ days of autonomy. The Asia Development Bank has financed over 200 solar mini-grid projects in Indonesia's eastern provinces since 2018. In the Pacific, the Australian Bushlight program pioneered the format — over 300 remote community systems installed since 2002 that replaced individual household diesel gensets with shared solar microgrids — and remains a reference design for load-limiting and demand management.

The battery and inverter choices that decide uptime

Component selection matters more in remote settings than in grid-tied work, because a truck roll can cost $2,000–$5,000 and take three weeks. LFP chemistry dominates for a structural reason: thermal runaway onset is roughly 200°C higher than nickel manganese cobalt (NMC), and remote sites frequently see 40°C+ ambient without active cooling. Cycle life at 80% depth of discharge runs 4,000–6,000 cycles for LFP versus 2,000–3,000 for NMC, which maps to 12–15 years of daily cycling against 6–8. Batteries on remote sites are typically 48 V or 400 V DC racks, need a BMS with per-cell monitoring, and — critically — need heaters or thermal enclosures if ambient drops below 0°C, as it does in high-altitude Andean and Himalayan villages.
Inverter selection follows three rules. First, surge capability of at least 2× continuous rating for 5–10 seconds, to handle motor starts. Second, a measurable grid-forming mode with droop control so multiple inverters can parallel without a communication bus — this matters because a single point of comms failure should never black out the village. Third, an integrated maximum power point tracking (MPPT) stage rated 10–20% above array nameplate so clipping losses stay under 1% on cold, clear mornings. A well-specified 400 V three-phase system with 96–98% conversion efficiency is achievable; anything under 94% is a red flag on component quality.

Sizing rules and the mistakes that kill projects

The most common technical failure is a load forecast built on survey answers rather than metered behaviour. Villages under-report evening demand, and productive-use loads — milling, welding, cold storage — arrive 12–24 months after energisation. Designers who size for today's load end up with a permanently saturated system and a queue of rejected connection applications.
A workable rule is to size the array at 1.5× the projected year-five peak demand in kWp, and storage at 1.2–1.5 days of average daily consumption in kWh, then verify against hourly simulation across a typical meteorological year. Diesel-downsizing hybrids should target a 25–35% renewable penetration floor minimum; below that, the genset runs at inefficient part-load and fuel savings evaporate. Above 70% penetration with lead-acid storage, cycle life collapses — which is one of the reasons LFP has displaced lead-acid in essentially every tender since 2021.
Distribution losses deserve a target too: keep three-phase runs under 500 metres where possible, use 16 mm² minimum service drops, and design for under 6% technical losses at peak. Beyond 800 metres, a second smaller satellite system normally beats a long feeder.

Tariffs, financing and the regulatory gap

Microgrids live or die on tariff design. Energy-based tariffs are fair but expose operators to weather risk; power-based tariffs (a fixed monthly fee scaled by the household's load limit) transfer demand risk and are easier for customers to predict. Many operators now run a hybrid: a small connection fee plus prepaid volumetric energy at $0.15–$0.25/kWh. Collection rates on prepaid systems hit 90–95%, against 60–70% on postpaid monthly billing in the same markets.
The regulatory piece is unfinished. Roughly 30 countries now have some form of mini-grid framework, but only a handful — Nigeria and Tanzania among them — specify what happens when the main grid arrives, standardising compensation or conversion of the microgrid's assets. Without that clause, lenders price in a political risk premium of 300–600 basis points, and projects that would clear at $0.22/kWh stall at $0.32/kWh. The World Bank estimates universal access by 2030 requires $30–40 billion per year in investment; current off-grid mini-grid flows run an order of magnitude below that.

What buyers and EPCs should specify in 2025

Three procurement trends have hardened into expectations over the last two years. First, metering and remote monitoring are no longer optional: a microgrid without per-customer telemetry cannot measure demand growth, cannot enforce load limits, and cannot finance itself. Second, modularity is priced in — most vendors now build in 20–50 kWp blocks so capacity can be added without replacing the inverters or the distribution backbone. Third, hybrid-ready design: pre-wiring the DC bus for a genset that is not yet installed, so a future productive-use customer can be served without a full redesign.
Component warranties have also stretched. Ten-year performance warranties on modules and 10-year/6,000-cycle warranties on LFP racks are now standard in donor-funded tenders, up from five years in 2019. For operators comparing full-life costs, that warranty coverage is worth more than a 3–5% capex saving on a cheaper module. DLXN's remote-site reference builds, documented in its project portfolio, follow the same modular pattern — a DC-coupled array, LFP storage sized for autonomy rather than peak, and grid-forming conversion — with the design margin concentrated in storage and distribution rather than in nominal array size.
The engineering problem of the remote microgrid is largely solved. What remains is the hard part: financing that survives a 15-year horizon, tariffs that a rural household can actually pay, and a maintenance model where a failed string in month 40 doesn't become a dead village in month 41.