Nepal's electricity supply is dominated by hydropower, while its economy-wide energy use remains dominated by biomass and commercial fuels. The transition challenge is therefore not only to add renewable generation, but to electrify end uses and coordinate energy across hours, seasons and sectors.

0.4
MWh per person per year, current electricity use in the 2050 modelling baseline
3.92 GW
DoED operating hydro capacity above 1 MW, a registry with defined scope, updated 16 July 2026
1,269 GWh
Technically feasible pumped-storage potential in a nationwide geospatial study
145–184 GWh
Projected second-life EV-battery capacity available by 2043

1. The clean-electricity paradox

DoED's operating registries list 3,915.744 MW of hydropower projects above 1 MW and 154.74 MW of solar projects, both updated 16 July 2026. Within those two registry categories, hydropower represents 96.2% of listed capacity. That percentage is a scoped registry calculation, not a complete census of every sub-1-MW, off-grid, thermal, biomass, wind or cogeneration facility. [10][11]

The reviewed 2050 paper reproduces an economy-wide energy balance totaling about 532.42 PJ: approximately 340 PJ of traditional biomass, 137 PJ of commercial fuels excluding grid electricity, 38.49 PJ of grid electricity and 16.5 PJ of modern renewables. Its Figure 1 caption labels the balance FY 2023/24, while the cited official WECS webpage is titled Energy Sector Synopsis Report 2024 (FY 2079/80). Because those period labels conflict, this review reports the numerical values but does not present either fiscal-year label as unambiguously resolved. The percentages below are author calculations from the four quantities. [1][8]

This distinction matters. A country can have a nearly renewable power grid while households still cook with fuelwood, vehicles burn imported petroleum and industry relies on coal or other thermal fuels. The transition therefore requires both clean supply and electrification of demand.

Figure 1 — Economy-wide energy-balance values reproduced in the 2050 study
Petajoules and calculated shares of the 532.42 PJ total. This is not an electricity-generation mix; the source-period label is disputed between the paper caption and official WECS webpage.
Biomass
340 PJ · 63.9%
Commercial fuels
137 PJ · 25.8%
Electricity
38.5 PJ · 7.2%
Modern renewables
16.5 PJ · 3.1%
Values reproduced by Bhatta and Lohani (2025), citing WECS. The paper caption says FY 2023/24; the linked official WECS page says Report 2024 (FY 2079/80). Shares are author calculations from the reported PJ values. [1][8] official report page ↗

The seasonal mismatch

RoR and peaking-RoR plants dominate Nepal's hydro fleet, so output is strongly coupled to river flow. The 2026 PHES study reports about 14 TWh of RoR and peaking-RoR generation in FY 2024/25, a wet season concentrated in June–September, and dry-season river discharge typically more than 50% below monsoon flows. The same paper cites wet-season exports of roughly 500–560 GWh per month and dry-season imports of roughly 170–360 GWh per month in the FY 2024/25 balance. [2][14]

Nepal's core planning problem is a time-shift problem: move abundant wet-season energy into dry months, and move midday solar into evening demand.

In the reviewed 2050 optimization, the ambitious-hydro, low-demand cases combine 170 TWh/year of modelled hydro generation with 75 TWh/year of demand and produce the highest LCOE values: $114.11/MWh without imports and $113.99/MWh with imports. The result is model-specific evidence that capacity expansion can be costly when annual output is poorly matched to demand and flexibility. [1]

2. Targets are expanding faster than the current system

Nepal's NDC 3.0, dated 14 May 2025 and submitted to the UNFCCC in May 2025, targets 14,031 MW of renewable electricity capacity by 2030 and 28,500 MW by 2035. Mini- and micro-hydro, solar, wind and bioenergy are targeted to supply 10% of total capacity by 2030 and 15% by 2035, compared with a 2024 reference share of 5%. [9]

The same NDC targets BEVs at 90% and 95% of private-passenger-vehicle sales in 2030 and 2035, and 70% and 90% of public-passenger-vehicle sales. It also calls for adequate charging infrastructure, strengthened battery-management and recycling practices, and pilot activity by 2030. [9]

Figure 2 — Renewable electricity capacity pathway in NDC 3.0
The colored bar shows total renewable capacity. The note below gives the targeted share from mini/micro hydro, solar, wind and bioenergy.
3.5
2024 reference
5% diversified sources
14.031
2030 target
10% diversified sources
28.5
2035 target
15% diversified sources
Capacity in GW. Source: Government of Nepal, NDC 3.0, electricity-generation and supply target table. [9] official PDF ↗

As of 16 July 2026, DoED listed 197 operating hydro projects above 1 MW totaling 3,915.744 MW and 27 operating solar projects totaling 154.74 MW. Ten additional solar projects totaling 74.6 MW held construction licenses. These are licensing-registry totals and should be cited with their category scope. [10][11][12]

Interpretation

Interpretation: the official targets broaden the supply mix, but the current operating registry remains hydro-centric. Reaching the NDC pathway requires faster non-hydro deployment, network reinforcement, flexible demand, and storage or trade arrangements that can manage both daily and seasonal variability.

3. Explore twelve least-cost pathways to 2050

Bhatta and Lohani evaluated twelve 2050 systems at hourly resolution. The factorial design combines three annual electricity-demand levels—2, 5 and 9 MWh per person—with two fixed hydropower pathways and either disabled or backup electricity imports. [1]

These are conditional optimization experiments, not forecasts or construction plans. They isolate how demand growth, hydropower build-out and regional trade alter the least-cost mix of PV, PHES and interconnection.

Table 1. Key assumptions that define the 2050 scenario results.
Model elementRepresentationInterpretive consequenceValidation
Demand2, 5 or 9 MWh/person/year; national annual demand of 75, 187 or 337 TWhScaling the same normalized hourly profile does not capture future end-use reshaping from EVs, cooking and industry.paper ↗
HydropowerReference pathway ≈13.27 GW in Table 1 (3.24 GW operating plus a 10.03 GW licensed-construction pipeline); ambitious pathway ≈37.545 GWHydro capacity is fixed within a pathway; rising demand is met by PV, PHES and imports.Table 1 ↗
ImportsBackup supply used only when domestic generation and storage cannot meet loadImport-enabled cases represent flexibility access, not unconstrained replacement of domestic generation.methods ↗
StoragePHES only in the least-cost modelBattery and other short-duration options are intentionally excluded; results cannot be read as proof that PHES is optimal for every grid service.methods ↗
Finance5% baseline discount rate; sensitivity from 3% to 10%Capital-intensive PV, hydro and PHES economics change materially with financing conditions.sensitivity ↗
Interactive Figure 3 — Nepal 2050 scenario explorer
Choose a demand level, hydropower pathway and trade condition. The bubble plot shows all twelve scenarios: x = solar PV capacity, y = PHES power, bubble size = annual demand, and label color = import condition.
A2-Import
Import enabledNo importBubble size = annual demand
All twelve values are transcribed from Table 4 of Bhatta and Lohani (2025). Click or focus a bubble to load that scenario. [1] paper + Table 4 ↗

What the scenario space shows

  • Solar scales with demand in most pathways. In A3-NoImport, 337 TWh/year of demand requires 345 GW of PV and 91 GW of PHES. Allowing backup imports reduces those capacities to 234 GW and 52 GW.
  • Hydropower expansion can be underutilized at low demand. The two B1 cases model 170 TWh/year of hydro generation against 75 TWh/year of demand and have the highest LCOE values in Table 4.
  • Trade substitutes for part of domestic balancing infrastructure. Across medium- and high-demand pairs, import access cuts modelled PHES power by about 43% to 79%; the exact reduction depends on the hydro pathway.
  • Financing is a first-order variable. For A2-Import, the paper reports LCOE rising from about $61/MWh at a 3% discount rate to about $84/MWh at 10%. The rounded endpoints imply a 37.7% increase; the paper describes the change as about 37%.

The A2-Import one-at-a-time sensitivity analysis uses a baseline LCOE of $66.89/MWh. A ±25% change in PV cost moves LCOE to roughly $61.8–72/MWh, while the broader discount-rate sweep from 3% to 10% moves LCOE from about $61/MWh to $84/MWh. These results support the paper's emphasis on PV procurement and financing, but they remain conditional on all other model inputs. [1]

4. A storage portfolio, not a single storage winner

Taken together, the reviewed studies support a portfolio interpretation rather than a universal technology ranking: PHES for bulk long-duration shifting; batteries for fast, modular and distributed services; second-life EV packs where state-of-health and safety can be managed; hydrogen where a molecule or long-duration sector link has value; and trade or flexible demand where they avoid physical overbuild. This is an editorial synthesis across studies with different scopes, not the output of one co-optimization model. [1][2][3][5]

Figure 4 — Conceptual energy-flow architecture for Nepal
Not to scale. The diagram distinguishes primary energy sources, balancing mechanisms and end uses.
Monsoon run-of-river hydropower
Solar PV: utility, rooftop and agrivoltaic
Reservoir hydro and regional imports
Pumped hydro
daily + seasonal shifting
Batteries and second-life EV packs
fast + distributed balancing
Electrolyzers + hydrogen
sector coupling
Exports, flexible loads and demand response
Households and electric cooking
Electric mobility and charging
Industry, heat and productive use
Dry-season supply and exports
Editorial design principle: route surplus energy to the highest-value feasible use while respecting storage losses, network constraints and end-use demand.
Author synthesis based on system roles evaluated in Refs. 1–6. The arrows are conceptual and not a quantified Sankey balance.

Distributed flexibility: the Thakle microgrid lesson

National planning models determine how generation, storage and trade can balance Nepal's power system, but distribution-level projects determine whether reliable electricity reaches remote communities. Shah et al. examined this lower level of the transition through a HOMER Pro analysis of Thakle Namuna Basti, a post-earthquake settlement in Melamchi Municipality comprising 32 households and approximately 130 residents. The field-survey-based load profile is transcribed from the paper's HOMER load-profile figure as 236.99 kWh/day with an approximately 41.5 kW peak; these two values are graph-derived rather than stated in machine-readable body text. The mean solar resource is explicitly stated as 4.97 kWh/m²/day. [16]

32
households modelled
237
kWh/day average load
41.5
kW peak demand
0.0588
$/kWh optimized LCOE

Under the paper's initial grid-connected assumptions, HOMER identified a lowest-net-present-cost architecture combining approximately 5.89 kW of micro-hydro, 10.9 kW of solar PV and an 8 kW converter with the utility grid. The reported optimized case had a net present cost of about $70,507 and a levelized cost of electricity of approximately $0.0588/kWh. The analysis also found that project economics improved when river flow and the electricity sell-back rate were higher, and deteriorated when purchased-power price or hydropower cost increased. [16]

Why this belongs in a national flexibility review. The Thakle study does not establish a national least-cost pathway. It demonstrates a complementary principle: site-specific combinations of micro-hydro, PV, storage, converters and grid exchange can reduce the cost and reliability burden at the distribution edge. Such systems can serve settlements before major transmission reinforcement arrives, while also creating controllable local generation and demand that may participate in a more flexible national grid.

The result should not be generalized without qualification. It depends on the surveyed load, assumed component costs, river-flow availability, grid tariff, sell-back price and HOMER's 25-year project representation. The paper also modelled lead-acid storage, so its battery conclusions should not be transferred directly to contemporary lithium-ion systems. The case is most useful as a reproducible planning template rather than a universal tariff benchmark. conference paper DOI ↗ full ML·Energy·Bio review ↗

5. Pumped hydro: from atlas points to field-validated sites

The ANU atlas is a topographic screening inventory, not a feasibility study. Dhakal et al. therefore applied district-boundary, land-cover and terrain filters in ArcGIS, verified reservoir boundaries and contours, cross-checked elevations with handheld GPS measurements whose typical open-sky vertical accuracy was stated as ±5 m, and conducted field visits. This workflow reduced 142 atlas reservoirs in Kavrepalanchwok to two physically suitable candidate pairs for prefeasibility analysis. [2][13]

Figure 5 — PHES screening and validation pipeline
The unit changes from individual atlas reservoirs to pairs and then candidate projects; the sequence represents decision gates rather than a uniform statistical funnel.
142
Atlas reservoirs initially identified in Kavrepalanchwok
62
PHES pairs fully inside district boundaries
35
Candidate pairs after terrain and land-cover filtering
3
Priority sites after geospatial assessment
2
Sites verified as physically suitable through GIS and field surveys

Banepa–Dhulikhel

~790 mHead difference
~3.1 kmReservoir separation

Roshi

~793 mHead difference
~2.7 kmReservoir separation
Source: Dhakal et al. (2026), abstract, Tables 1–3 and field-validation sections. The count changes from reservoirs to pairs, so this is a sequence of decision gates rather than a statistical attrition rate. [2] paper ↗

The economic result depends on charging electricity

For the field-validated Roshi configuration, the study used a 793 m head and 1.3 hm³ active reservoir volume, estimating 720 GWh/year discharged and 888 GWh/year pumped. The $238.5 million initial capital estimate includes construction, 20% risk and contingency, transmission, 10% miscellaneous cost, and 2.5% design and consultancy. LCOS is $105/MWh with RoR PPA-priced charging, $99/MWh with solar-PV charging, and $65/MWh with subsidized charging. [2]

Figure 6 — Roshi PHES economics and capital-cost structure
LCOS under three pumping-cost assumptions; capital-cost shares are the study's reported component distribution.
$105
RoR PPA charging
$99
Solar PPA charging
$65
Subsidized charging
Levelized cost of storage, $/MWh

Initial capital cost: $238.5 million

18%13%42%26%
Civil · 18%Hydromechanical · 13%Electromechanical · 42%Electrical infrastructure · 1%Development + contingency · 26%

Interpretation: because charging electricity is an LCOS input, tariff design and access to low-value surplus energy materially affect the storage case. Capacity, ancillary-service and avoided-curtailment revenue would require a separate market-value analysis.

Source: Dhakal et al. (2026), Tables 3–4 and LCOS scenarios. Capital shares are grouped from the study's itemized ICC table; rounding may not sum exactly. [2] paper ↗

Nationwide potential remains much larger than modelled need

A separate national reservoir-pairing study estimated 3,012 GWh of theoretical PSH potential and 1,269 GWh of technically feasible potential. Its flat-land-to-river configuration contributed 2,716 GWh theoretical, 1,198.8 GWh technically feasible and 897.9 GWh exploitable under the study's grid, infrastructure and environmental filters. [4]

The 1,269 GWh result and the approximately 50 TWh reported for the Global Greenfield Atlas are not competing estimates: they use different reservoir configurations, elevation limits, pairing rules and screening criteria. The robust conclusion is that mapping must be followed by site-specific geological, environmental, social, network and financial assessment.

6. Second-life EV batteries: transport becomes grid infrastructure

The 2026 EV-battery study defines end of first vehicle life at 80% of nominal capacity and models LFP, LMO and NMC degradation under assumed Nepal usage and environmental conditions. Estimated first-life durations are about 7.0 years for LFP, 3.3 years for LMO and 3.7 years for NMC. Those timings determine when packs enter a potential second-life stock. [3]

Across EV-growth and chemistry-mix scenarios, cumulative second-life capacity reaches 145–184 GWh by 2043. For 2040, the study estimates 67% to more than 94% coverage of average daily wet-season surplus in its higher-surplus S-1 cases; in lower-surplus S-2 cases, projected second-life capacity exceeds the average daily surplus. These are energy-capacity comparisons, not demonstrations of seasonal firming, deliverable power, network access or economic dispatch. [3]

Interactive Figure 7 — Battery chemistry and retirement timing
Typical chemistry inputs used in the study. First-life duration is the modelled time to 20% capacity loss under assumed Nepal operating conditions.
7.0years first life

Lithium iron phosphate (LFP)

The study reports ranges because EV uptake, chemistry shares, degradation and hydropower surplus vary by scenario. Input values are transcribed to the review data file. [3] paper ↗
Figure 8 — Share of average daily 2040 wet-season surplus that second-life batteries could absorb
Published scenario ranges. S-1 represents higher projected surplus; S-2 represents lower projected surplus.
Higher-surplus S-1
67–94%
Lower-surplus S-2
>100%
2043 storage stock
145–184 GWh
The paper reports ranges rather than a single deterministic outcome because EV uptake, battery chemistry and hydropower surplus all vary by scenario.[3]

The LFP timing paradox

LFP has the longest modelled first-life duration of the three chemistries. A higher LFP market share therefore delays the arrival of retired packs within the 2040 planning window and slightly reduces near-term second-life stock—not because LFP performs worse, but because it remains above the vehicle-retirement threshold longer.

The practical storage resource will also depend on pack collection, traceability, state-of-health diagnostics, repurposing design, inverter compatibility, thermal management, fire codes, warranties and end-of-second-life recycling. The national energy-capacity estimate should therefore be treated as an upper-level resource assessment, not deployable capacity.

7. Hydrogen is valuable when it connects sectors

The 2022 study represents Nepal with 52 nodes, 68 transmission lines and 8,760 hourly time steps. It estimates about 32 GW of installed generation for domestic electricity demand by 2050 and about 14 GW of additional hydropower when road transport is fully electrified in the model. [5]

For the export-oriented scenario with equal battery-electric and hydrogen-vehicle shares, the paper's abstract reports 5.7 GW of electrolyzers, 12 GW of hydrogen-storage-tank capacity and 23 GW of storage-based hydropower. The body text separately describes export-oriented hydro-storage deployment as “about 27 GW.” Because the source is internally inconsistent, this review does not treat 23 GW or 27 GW as an independently resolved exact value. The robust result is that large storage-hydro deployment occurs mainly in the export-oriented case, while hydrogen-system deployment is negligible in scenarios without hydrogen vehicles under the paper's cost assumptions. [5]

A 2021 Nepal perspective reports electricity–hydrogen–electricity round-trip efficiency of roughly 20–30%, compared with about 80–90% for batteries or PHES. This supports selective use of hydrogen where long-duration storage or the hydrogen molecule itself provides sector value. The examples of fertilizer, industrial feedstock, high-temperature heat and heavy transport are editorial applications, not outputs of the 2022 power-grid model. [5][6]

Table 2 — Functional roles in a diversified flexibility portfolio
Technologies should be compared by the system service they provide, not by a single cost or efficiency metric.
OptionBest-fit timescalePrimary valueNepal-specific evidenceMain constraint
Closed-loop PHESHours to days; potentially seasonalBulk energy shifting, reserve, frequency response, long asset lifeTwo field-validated Kavrepalanchwok sites; Roshi LCOS $65–105/MWhHigh initial capital, site validation, land and transmission
Second-life EV batteriesSeconds to hours; distributed daily cyclingFast response, local congestion relief, circular use of imported batteries145–184 GWh projected by 2043; 67–94% of S-1 daily surplus in 2040State-of-health uncertainty, aggregation, fire safety, warranties
New battery systemsSeconds to sub-day balancingFrequency control, peak shaving, solar time shifting, weak-grid supportThe PHES paper cites battery cost escalation beyond roughly 4–6 h; chemistry inputs in the EV study exceed 95% for charge and discharge steps.Replacement cycles, import dependence, recycling
Green hydrogenLong duration and sector couplingFuel and feedstock for transport and industry; curtailment conversionHourly grid model shows reduced RoR/PROR curtailment and 5.7 GW electrolyzers in an export-oriented scenarioLow power-to-power efficiency and high infrastructure cost
Regional trade + flexible demandReal-time to seasonalShares diversity across geography; avoids some storage and curtailmentIn medium/high 2050 scenario pairs, import access reduces PHES power by about 43–79%, depending on hydro pathway.Market rules, transmission, geopolitical and contractual risk
Cross-study synthesis. Timescales and constraints are functional descriptors; Nepal-specific numerical claims link to the underlying studies and audit ledger.

8. An evidence-informed 2026–2035 roadmap

The following sequence is the author's synthesis of the evidence and identified implementation gaps. It is not an adopted Government of Nepal plan and should be evaluated through stakeholder consultation, system studies and project-level safeguards.

Correct the planning objective

Optimize for reliable energy delivered across seasons and sectors, not simply installed megawatts. Publish hourly and seasonal supply-demand data, curtailment and reservoir operating information.

Accelerate solar beyond a token share

Use utility solar, rooftops, canal and reservoir surfaces, brownfields and agrivoltaics. The 2050 model suggests solar may provide roughly 40–60% of generation in deeply electrified pathways, well above current policy shares.

Move PHES from atlas screening to bankable pilots

Complete geotechnical, hydrological, environmental and social studies for priority closed-loop sites. Establish market products for capacity, ancillary services and avoided curtailment — not only energy arbitrage.

Create a second-life battery regime before mass retirement

Require digital battery passports, health diagnostics, repurposing certification, fire codes, installer training, extended producer responsibility and end-of-second-life recycling.

Use hydrogen selectively

Prioritize pilots where hydrogen displaces imported fossil feedstock or fuels, rather than treating it as the default electricity-storage medium. Co-locate electrolyzers with constrained renewable generation and industrial demand.

Strengthen regional trade without outsourcing security

Expand transmission and transparent trading arrangements while maintaining sufficient domestic storage and reserve. Trade should reduce cost, not create a new single-point dependency.

De-risk capital

Use concessional finance, guarantees, standardized PPAs and competitive procurement. The least-cost model is highly sensitive to discount rate; financing reform can change the optimal system as much as technology cost.

Grow productive electricity demand

Coordinate EVs, electric cooking, heat pumps, industrial furnaces, cold chains and irrigation with renewable availability. Flexible demand is an asset when tariffs reward consumption during surplus periods.

9. What the evidence does — and does not — establish

Scenario, not forecast

The 2050 capacity results depend on demand, cost, climate and trade assumptions. They show system relationships, not a precise construction schedule.

Atlas, not feasibility study

Most mapped PHES sites have not undergone geology, land-tenure, biodiversity, heritage or community-consent assessment.

Battery stock is policy-dependent

The 145–184 GWh estimate assumes EV adoption, charging infrastructure and battery-retirement trajectories broadly follow the modelled pathways.

Historical hydro is not future hydro

Climate change, sediment, glacier retreat and extreme events can alter river flows and asset reliability. Future models need climate-adjusted hydrology.

Aggregated demand hides grid constraints

National energy balance does not guarantee local voltage, frequency, inertia or transmission adequacy. Distribution planning remains essential.

Costs are context-sensitive

LCOS and LCOE depend on financing, construction risk, utilization, tariff design and exchange rates. Published values are comparable indicators, not bids.

Conclusion

The reviewed resource assessments and conditional scenarios indicate that a deeply renewable Nepalese energy system is technically plausible, but they do not establish a single bankable national build plan. The implementation problem is coordination: demand growth, tariff design, site validation, transmission, financing, environmental and social safeguards, and governance of new assets such as repurposed EV batteries.

Across the studies, the recurring architecture is a complementary solar–hydro system with multiple flexibility layers. PHES is evaluated for bulk shifting, batteries for modular and distributed services, regional trade for geographic balancing, and hydrogen for selected sector-coupling cases. Their relative scale must be resolved by models that co-optimize these options under updated demand, hydrology, network and cost data.

The strategic shift is from counting renewable megawatts to designing a flexible renewable-energy system.
Verification method. This revision separates source facts, calculations, model outputs, cross-study synthesis and editorial recommendations. Claims were checked through independent primary-source, arithmetic, official-current-data, citation-metadata and code-validation passes. Three qualifications remain visible rather than being concealed: the WECS fiscal-year label conflict, two graph-transcribed Thakle load values and the hydrogen paper's 23-versus-27 GW internal discrepancy. Numerical graphics were rebuilt from reported values; no publisher figure was copied.
Cite this review
Upreti, Saugat. “From Monsoon Surplus to a Flexible Grid: An Evidence-Audited Review of Nepal’s Renewable-Energy Transition.” ML·Energy·Bio, July 2026. https://mlenergybio.com/blog/nepal-flexible-grid.
@misc{mlenergybio2026nepalflexiblegrid,
  author = {Upreti, Saugat},
  title  = {From Monsoon Surplus to a Flexible Grid: An Evidence-Audited Review of Nepal's Renewable-Energy Transition},
  year   = {2026},
  month  = {July},
  url    = {https://mlenergybio.com/blog/nepal-flexible-grid},
  note   = {Evidence-audited review with interactive scenario analysis}
}