Off-Grid Microgrid Feasibility Study: Step-by-Step Guide and Checklist

Key takeaways

• Define the decision, site boundary, and reliability rule before searching for equipment sizes; otherwise the model can answer the wrong question precisely.
• Use an 8,760-hour load profile for a serious feasibility decision, preserve chronology and timestamps, and document any synthetic or scaled intervals.
• Review difficult dispatch hours, state-of-charge limits, generator operation, curtailment, fuel, and reserve behavior before accepting the lowest-cost feasible case.
• Sensitivity analysis is part of feasibility, not decoration: fuel delivery, load growth, resource uncertainty, component cost, replacement timing, and reliability rules can change the recommendation.
• Hand off a small set of defensible cases and unresolved risks to the EPC or detailed engineer; do not present a screening optimum as a construction design.

What an off-grid microgrid feasibility study should decide

A useful feasibility study is a go, no-go, or gather-more-data decision package. It should say whether an off-grid PV, battery, and diesel concept can serve the agreed load under an explicit reliability rule, what the credible capacity range costs, which assumptions control the answer, and what work must happen before procurement.

That is broader than finding one attractive size. The IEA PVPS feasibility blueprint organizes the work around scope, data, modeling, and recommendations across technical, financial, and organizational questions. The U.S. Department of Energy project-development checklist similarly treats feasibility as one part of a staged path through planning, design, procurement, implementation, and operation.

Sandia describes conceptual microgrid design as an early level of development, roughly 10% to 20%, used to communicate goals and constraints before detailed engineers, vendors, and contractors finish the design. That is the right posture for this guide: enough evidence to make the next decision, with an honest boundary around what has not been engineered yet.

• Decision: proceed, pause, reject, or collect better data.
• Feasibility: capacity combinations that meet the stated energy and reliability constraints.
• Economics: lifecycle cost and operating exposure under documented assumptions.
• Risk: the inputs and events most likely to change the decision.
• Handoff: the cases, time-series evidence, and open questions the next engineer needs.

Set the study boundary before opening the model

Start with a one-page basis-of-study note. Name the approving decision, site, loads inside and outside the boundary, required operating modes, study period, currency, escalation assumptions, and the party responsible for each input. If the system must support only critical loads during certain conditions, define those loads and conditions explicitly.

Reliability needs a model-ready rule. Examples include zero modeled unmet energy in the study year, a maximum annual unmet-load fraction, operating-reserve requirements, a minimum battery state of charge, or survival through a specified outage or low-resource event. These rules are not interchangeable. A model can satisfy an annual energy target and still fail the site during one critical hour.

DOE guidance supports increasing data resolution as a project moves from screening to validation: annual information may support early identification, while hourly or subhourly time series are more appropriate for detailed analysis. For an off-grid sizing recommendation, chronology is usually central because storage and generator constraints depend on the order of the hours.

• Define whether auxiliary loads, conversion losses, black-start loads, and future phases are inside the load boundary.
• Separate normal load, critical load, deferrable load, and optional productive load where the distinction matters.
• State whether the diesel generator is emergency-only, routinely dispatched, or restricted by run hours, minimum load, emissions, or fuel storage.
• Record what the model does not represent, such as feeder power flow, protection, harmonics, transient stability, controls implementation, civil works, or permitting.

The 12-step feasibility workflow

The sequence below keeps the analysis tied to a real approval gate. A smaller study can compress the documentation, but it should not silently skip the decisions represented by these steps.

1. Write the decision and approval gate. State who will use the study and what action a positive result authorizes.
2. Define the site and system boundary. List included loads, operating modes, future phases, and exclusions.
3. Set reliability and unmet-load rules. Translate service expectations into explicit constraints and review metrics.
4. Build and validate the 8,760-hour load. Preserve timestamps, units, time zone, gaps, seasonal shape, and the relationship between annual energy and peak demand.
5. Select and document the solar resource. Record the location, year or typical-year method, plane-of-array assumptions, losses, and whether the data are measured, satellite-derived, or synthetic.
6. Specify PV, battery, converter, and generator constraints. Include efficiencies, usable state-of-charge range, power limits, degradation or replacement assumptions, generator loading, and availability.
7. Build delivered-fuel and lifecycle-cost assumptions. Include transport, storage losses, escalation, capital cost, replacements, fixed and variable O&M, project life, discounting, taxes, and incentives only when applicable.
8. Establish a diesel-only or current-supply baseline. Apply the same load, project life, fuel boundary, and economic conventions used for the hybrid alternatives.
9. Search feasible capacity combinations. Reject cases that violate hard rules before ranking the survivors by the chosen objective.
10. Audit difficult hours and binding constraints. Inspect low-state-of-charge periods, generator starts and loading, unmet load, reserve shortfall, curtailment, and energy-balance closure.
11. Run decision-changing sensitivities. Vary uncertain inputs far enough to expose a different recommendation, not merely to produce a decorative tornado chart.
12. Package the recommendation, alternatives, risks, and handoff. Keep the preferred case, a lower-capital or boundary case, and at least one reliability or growth case available for the next stage.

Microgrid feasibility study input checklist

Every important input needs a value, unit, source, owner, date, confidence level, and planned update. The minimum below is deliberately practical: it is enough to expose the assumptions that most often move an off-grid PV, battery, and diesel screen.

Worked example: a remote clinic under three delivered-fuel prices

To make the process concrete, we reran the current synthetic Remote clinic (small) benchmark on July 18, 2026. The load contains 8,760 hourly values totaling 97,887 kWh per year with a 20 kW peak. The synthetic tropical solar series produces 1,229.7 kWh per installed kWp before capacity search. The economic case uses a 25-year life, 8% nominal discount rate, 2% inflation, current model cost defaults, and no tax credit or other incentive.

The hard reliability constraint was 0% modeled unmet load. We changed only the delivered diesel price, then used the same deterministic coarse-and-refined search. The table reports the lowest-NPC candidate found within that sampled search space, not a claim of a continuous global optimum or a vendor-ready design.

• The low-price run evaluated 326 candidates and the other runs evaluated 325; each retained 102 candidates that passed the hard feasibility checks.
• The example uses synthetic demand and solar data. Replace both with site evidence before treating the result as a project recommendation.
• The cost defaults illustrate a reproducible method; they are not supplier quotations, installed EPC prices, or a financing offer.

How to interpret the worked example

As delivered fuel becomes more expensive, the selected design buys more solar capacity and burns less diesel. Annual fuel falls from about 1,941 liters to 432 liters across the tested range. The middle case also selects more battery energy than either endpoint, a useful reminder that capacity search is not a simple rule such as more expensive fuel always means a larger battery.

Total lifecycle cost still rises with fuel price. At $1.45 per liter, the selected design produces about 132,805 kWh of first-year PV energy and curtails or leaves excess about 33,408 kWh. That is not automatically a modeling error. It is a prompt to test whether water pumping, ice making, charging, thermal loads, or other productive demand can use surplus without creating a new critical-load obligation.

The fuel sensitivity changes both the capacity recommendation and operating exposure, so it belongs in the decision. A feasibility report that presents only the middle price would hide that relationship. It should also preserve other near-feasible or near-optimal candidates because a vendor constraint, land limit, battery warranty, or operator preference may make a neighboring design more practical.

Audit the difficult hours before accepting the economics

Annual totals are the start of review, not the end. For the $1.45-per-liter case, the first-year trace reaches a minimum battery state of charge of about 20.1%, records 245 generator operating hours and 23 starts, and completes about 199 equivalent battery cycles. The current run reports 0 kWh unmet load and a small annual reserve shortfall of about 0.056%, within the engine's current 2% reserve-deficit tolerance.

Those summary metrics tell the reviewer where to look. Export the dispatch, find the lowest-state-of-charge and largest-load hours, inspect the preceding multi-hour sequence, and confirm that generator power, battery limits, conversion losses, and reserve treatment behave as intended. Then rerun any assumption that appears to bind the answer.

• Energy balance closes for every hour and annually after losses and curtailment.
• State of charge never crosses the modeled minimum or maximum.
• Battery charge and discharge power remain within the selected converter and battery limits.
• Generator loading, run hours, starts, fuel rate, and derating are operationally credible.
• Unmet load and reserve shortfall are reported separately and checked against their exact acceptance rules.
• The worst sequences remain feasible after load growth, lower solar output, or equipment unavailability is introduced.

Run sensitivities that can change the decision

A good sensitivity plan starts with uncertainty and consequence. Use ranges that a source or stakeholder can defend, and state what decision would change at each edge. Do not vary everything blindly: prioritize inputs that can change feasibility, technology size, lifecycle cost, or the choice to proceed.

For uncertain critical load, NLR guidance recommends testing more than one percentage when the value is not known. The same principle applies to fuel delivery and load growth: preserve a central case, then test a credible low and high boundary. Weather uncertainty deserves chronological cases where possible because a flat annual derate does not reproduce a run of cloudy days.

• Fuel: delivered price, escalation, delivery interruption, minimum shipment, storage loss, and maximum on-site inventory.
• Load: data-quality correction, critical-load fraction, productive demand, start-up power, seasonal peaks, and phased growth.
• Resource: alternative weather years, soiling, temperature, array availability, orientation, and a prolonged low-solar sequence.
• Battery: installed cost, usable depth, round-trip efficiency, power-to-energy ratio, calendar and cycle degradation, and replacement year.
• Generator: efficiency curve, minimum loading, site derating, maintenance outage, overhaul cost, start limits, and available unit sizes.
• Economics: project life, real or nominal discount convention, inflation, capital contingency, installation cost, and incentive eligibility.
• Reliability: unmet-load threshold, reserve rule, minimum state of charge, equipment outage, and any required autonomy interval.

What the feasibility report should contain

The report should let a second analyst reproduce the conclusion and let a decision-maker see the unresolved risk. A glossy result page without the basis of study, input provenance, difficult-hour evidence, and sensitivity boundaries is not enough.

Know where feasibility ends and detailed engineering begins

A feasibility model can select credible capacity ranges and expose hard operating hours. It does not by itself specify the one-line diagram, conductor and transformer ratings, protection coordination, grounding, controls architecture, communications, fault duties, power quality, dynamic stability, civil design, fire safety, permitting, or commissioning plan.

Give the next engineer the actual interval data, model inputs and version, candidate sizes, hourly dispatch exports, difficult-hour timestamps, sensitivity cases, unresolved risks, and the reason each case was retained. The detailed design team can then test the real network and vendor equipment without reverse-engineering a single summary number.

• Carry at least the preferred, boundary, and growth or reliability cases into vendor and electrical review.
• Replace generic performance and cost assumptions with site- and vendor-specific evidence as it becomes available.
• Reconcile any network, protection, controls, auxiliary-load, derating, or equipment-availability constraint back into the energy model.
• Rerun the feasibility case after material design changes so the decision package and detailed design do not drift apart.

Bottom line

An off-grid microgrid feasibility study is complete enough to support a planning decision when the scope and reliability rule are explicit, the chronological inputs are traceable, at least one capacity combination passes the hard constraints, difficult hours have been audited, and the recommendation remains credible across the agreed sensitivities.

Use MicrogridModeler for the focused browser-first PV, battery, and diesel feasibility layer when deterministic results and an auditable handoff matter. Keep spreadsheets for input QA and independent commercial checks. Bring in broader optimization, circuit, protection, controls, vendor, civil, and permitting tools when the question crosses the screening boundary.

Keep exploring

Sources and review notes

This article is grounded in the cited technical sources and the stated modeling assumptions. Recheck project inputs, equipment data, and local requirements before using it for design.

FAQ