LIT_083: Mohseni, Brent, Burmester & Browne (2021) — Lévy-flight Moth-Flame Optimisation Micro-Grid Equipment Sizing…

LIT_083: Mohseni, Brent, Burmester & Browne (2021) — Lévy-flight Moth-Flame Optimisation Micro-Grid Equipment Sizing…

Source

doi:10.1016/j.egyai.2021.100047 — original publication (opens in a new tab; the file is not redistributed)

Summary

The methods companion to LIT_032: a peer-reviewed techno-economic optimisation of the Totarabank eco-village micro-grid (central Wairarapa, NZ), but built around a new global-search sizing algorithm — the Lévy-flight moth-flame optimisation algorithm (LF-MFOA) — with a nested linear-programming (LP) day-ahead energy-dispatch optimiser running over a moving 24-hour horizon. The authors argue that the full-factorial (HOMER) and linearised-exact (MILP) tools dominant in the micro-grid literature return sub-optimal designs because they cannot preserve the problem’s nonlinearity and non-convexity; a nature-inspired meta-heuristic can. Applied to the same 8-lot / 14-inhabitant grid-tied community as LIT_032, the LF-MFOA finds a cost-optimal mix of 17.5 kW solar PV, 30 kW wind, 41 kWh Li-ion battery and 9 kW inverter that fills the site’s existing transformer capacity, yielding a negative LCOE (−0.02/kWh) — i.e. lifetime grid-export and local-sale revenues exceed lifetime costs — and an ~18% higher project net present value than HOMER Pro's battery-less optimum. The LF-MFOA statistically outperforms nine established meta-heuristics (by ≥6.5% on whole-life cost), and the nested day-ahead dispatch adds ≥11.5% cost savings. The paper's value to Neobiome is a second, larger-system NZ community-scale techno-economic point and a documented, reproducible sizing method — not a fresh LCOE provenance for the 0.094/kWh figure, which remains LIT_032’s.

Key claims

- claim: "Test case is a grid-connected, DC-coupled community micro-grid for Totarabank, an eight-lot residential subdivision (permaculture-designed eco-community) with 14 inhabitants in the Wairarapa District, North Island, New Zealand (GPS 41°1′4″S 175°40′0″E); the site currently has an installed solar PV capacity of 11.4 kWp."
  source_location: "Section 3 (p.9); Fig. 9 (p.9); Section 4.5.4 (p.20)"
- claim: "Method: a Lévy-flight moth-flame optimisation algorithm (LF-MFOA) sizes the MG equipment (outer capacity-planning loop), with an intelligent LP-based day-ahead energy-dispatch optimisation (solved via MATLAB 'linprog' over a moving 24-h horizon) nested inside; sizing is subject to LPSP_max = 0 (load always satisfied) and a minimum self-sufficiency ratio SSR_min = 80%, over an 8,760-h annual operating simulation."
  source_location: "Abstract; Section 2 (pp.5–8); Section 2.2 (p.6); Section 2.3 (p.6)"
- claim: "Cost-optimal system (LF-MFOA proposed model): PV 17.5 kW, wind 30 kW, battery 41 kWh, system inverter 9 kW; total net present cost of the MG = −$50,332; LCOE = −0.020 $/kWh. Benchmark HOMER Pro optimum: PV 7.5 kW, wind 40 kW, battery 0 kWh, inverter 15 kW; TNPC = −$42,646; LCOE = −0.018 $/kWh. (A negative TNPC/LCOE means projected lifetime revenues exceed costs.)"
  source_location: "Table 4 (p.12)"
- claim: "The meta-heuristic (LF-MFOA) optimum indicated an increase of approximately 18% in the total net present value of the project relative to HOMER Pro's optimum — i.e. HOMER Pro underestimates the project's total net present worth by ~18%; profitability is improved by at least 18% for the community MG scheme."
  source_location: "Section 4.1 (p.12); Conclusions item 1 (p.21); Abstract highlights"
- claim: "Capital-budgeting metrics — proposed PV/WT/battery MG: LCOE −$0.02/kWh, modified internal rate of return (MIRR) 5.4%, discounted profitability index (DPI) 1.43. Existing PV-only baseline: LCOE $0.39/kWh, MIRR 2.2%, DPI 1.09. (Reinvestment rate assumed 0% in the MIRR.)"
  source_location: "Table 8 (p.20); Section 4.5.4 (p.20)"
- claim: "Cost decomposition (proposed system): total equipment-related cost $123,012 (total capital cost $109,172; total replacement cost $28,875; total O&M cost $6,621; total salvage value −$21,656) and power-exchange-related cost −$173,344; grid energy trading is net profitable at ~$2,517/yr; the entire $123,012 investment is recouped within ~10 years."
  source_location: "Section 4.4 (p.19)"
- claim: "Tariffs: existing feed-in tariff for excess renewable export = $0.08/kWh; grid imports / local sales priced at a flat retailer tariff of $0.23/kWh, described as the most recent average domestic electricity price at the studied site (ref [102]). All monetary values are 2019 NZ\$."
  source_location: "Section 4.5.4 (p.20); Section 3 (p.11)"
- claim: "Self-sufficiency: the optimal system's self-sufficiency ratio equals 80% (the minimum allowed), i.e. 20% of the yearly load is met by grid imports. Sensitivity over SSR_min from 0% to 100% (11 cases): forcing SSR_min = 100% increases MG lifetime cost by only 14% (≈ $8k) versus no constraint; SSR_min from 0%–60% yields the same least-cost solution (the unconstrained optimum already returns ~62% self-sufficiency)."
  source_location: "Section 4.3 (p.17); Section 4.6 (pp.20–21); Conclusions (p.22)"
- claim: "Annual on-site renewable generation, on average: ~20,382 kWh (~21%) solar PV and ~78,891 kWh (~79%) wind. Optimal generation-capacity mix = 63% wind / 37% solar PV (proposed model) versus 84% wind / 16% solar PV (HOMER Pro). Grid-outage survivability = 100% and battery-bank autonomy = 14 h for the proposed system (0 h for HOMER Pro's battery-less optimum)."
  source_location: "Section 4.3 (p.17); Section 4.1 and Table 4 (p.12)"
- claim: "Component techno-economics (2019 NZ\$): PV panel 375 W Canadian Solar KuMax $437/unit, 18.9% eff, 20-yr life; wind turbine 5 kW AWS HCM $6,450/unit, 20-yr life (cut-in 2.7 m/s, rated 11 m/s, cut-out 25 m/s); Li-ion battery pack 1 kWh generic $885/kWh (replacement $417/kWh, O&M $2.1/kWh/yr, 15 yr or 12,000 cycles, 92% charge/discharge eff); hybrid inverter 3 kW Selectronic SPMC240 $4,600/unit, 96% eff. Real interest rate 2.45%, project lifetime 20 years."
  source_location: "Table 3 (p.10); Section 3 (p.9)"
- claim: "The LF-MFOA returns a lower whole-life cost than the original MFOA, PSO, GA, hybrid GA-PSO, ALO, ABC, SA, HS and ACO — by at least 6.5%, 7.3%, 7.7%, 8.4%, 9.1%, 11.8%, 12.1%, 12.6% and 12.9% respectively (best of 30 trials); nesting the day-ahead operational optimisation adds cost savings of at least 11.5%."
  source_location: "Conclusions items 3–4 (pp.21–22); Abstract"
- claim: "Data provenance: forecast meteorological and wholesale-price profiles are 2010–2019 averages from NIWA and the NZ Electricity Authority; the household load profile is synthesised from the NZ GREEN Grid project's future household-demand estimates scaled to the site's population; low-temperature heat is assumed to be 46% of total household electricity use (27% space heating, 19% water heating)."
  source_location: "Section 3 / 3.2 (pp.10–11)"

Neobiome Intelligence relevance

• D01 Renewable Energy & Storage — a second, independent NZ community-scale sizing result for the same site as LIT_032, reached by a different (LF-MFOA) method: cost-optimal 17.5 kW PV + 30 kW WT + 41 kWh battery + 9 kW inverter, sized to the site’s transformer limit LIT_083 (Table 4). It supplies a transparent, reproducible sizing-and-dispatch methodology (nested LP day-ahead scheduling within a meta-heuristic capacity search) and a full 2019-NZ$ component-cost table (PV 437/375 W unit; WT 6,450/5 kW unit; battery 885/kWh; inverter 4,600/3 kW unit) usable as a cost-basis cross-check LIT_083 (Table 3). It also documents a methodological caution for any NI cost engine: HOMER-style tools underestimated this project’s net present worth by ~18% LIT_083 (Section 4.1).
• I01 Financial & Economic Sufficiency — a community-ownership investment case: MIRR 5.4% and DPI 1.43 for the optimised MG versus 2.2% / 1.09 for the existing PV-only setup, with grid trading net-profitable at ~2,517/yr and the 123,012 outlay recouped in ~10 years LIT_083 (Table 8, Section 4.4). The negative LCOE (−$0.02/kWh) means projected lifetime export/sale revenues exceed lifetime costs — the paper frames the system as a low-risk, high-return opportunity suitable for community ownership or third-party (PPA/lease) finance LIT_083 (Section 4.5.4).
• I06 Resistance to External Shocks — quantifies energy resilience for the optimised system: grid-outage survivability 100% and a battery-bank autonomy of 14 h, against 0 h for HOMER Pro’s cost-optimal battery-less design LIT_083 (Table 4). Demonstrates that including a battery for arbitrage also delivers a resilience by-product at this scale.
• I07 Fulfilment of Basic Needs — a directly reusable self-sufficiency-economics result: the design meets an 80% self-sufficiency target, and pushing that to 100% costs only ~14% more over the life-cycle (≈ $8k), because the site has already surpassed grid parity LIT_083 (Section 4.6). This “relative cost of self-sufficiency” curve is a template for the SSR-vs-cost trade-off an NI energy design must navigate — though it is specific to a 14-inhabitant, high-wind, grid-tied site and needs scaling before use in a 50+ household worked example.

Notes

• Not a duplicate of LIT_032; it completes the re-scoped RT_081. Both papers (same authors, same Totarabank site) are separate studies with different objective functions, horizons and solvers, so they report different optima and different LCOEs by design. LIT_032 (2020 Energies) = resilience-constrained expansion of the existing 11.4 kW PV → LCOE 0.094/kWh at retail 0.34/kWh. This paper (2021 Energy and AI) = full LF-MFOA re-sizing to the transformer limit → LCOE −0.02/kWh at retail 0.23/kWh. This ingest confirms LIT_032’s correction that the $0.094 LCOE originates in the 2020 paper, not this one.
• Retail-price discrepancy to weigh at calculation time. The site retail tariff here is 0.23/kWh (2019 NZ\, ref [102], “most recent average domestic electricity price at the studied site”); LIT_032 uses $0.34/kWh (“current average domestic electricity price in the Wairarapa region”). Different scopes (studied-site vs regional) and citations, from the same team — do not treat as a single agreed figure. The NZ retail-price recalibration gap already logged on LIT_032 covers both.
• Modelled optimisation outputs, not measured outturns; deterministic parametric inputs (the authors flag probabilistic uncertainty as future work). Network charges, service fees and transformer-upgrade costs are excluded, and grid exchange is bounded by the existing transformer — so the negative LCOE understates true system cost if grid-parity were priced with those adders (the authors say so explicitly, Section 4.6 / Conclusions).

Research targets

Documents to retrieve

• RT_081 (this source) — RESOLVED. Mohseni, Brent, Burmester & Browne (2021), Energy and AI 3:100047, the Lévy-flight MFOA MG-sizing paper the re-scoped target sought for method comparison. Ingested as LIT_083. Confirms the LF-MFOA sizing method and supplies a second NZ community techno-economic point for Totarabank; it is NOT the provenance of the $0.094/kWh LCOE (that remains LIT_032).

Research gaps

• No new RTs. The residuals this paper surfaces are already tracked on LIT_032: NZ retail-price recalibration (the 0.23 vs 0.34/kWh spread sits inside it), 2019-NZ$ component-cost vintage, and the scale gap from 14 inhabitants to a 50+ household community. The authors’ own future-work items (large-scale systems, probabilistic uncertainty, multi-objective meta-heuristic variants, grid-parity with network charges) are methodological, not NI data gaps.

Connections

Links to

Sources (1): LIT_032

Referenced by

SSI indicators (3): I01: Financial & Economic Self-Sufficiency · I06: Resistance to External Shocks · I07: Fulfilment of Basic Needs

EDT domains (1): D01: Renewable Energy & Storage Systems

Sources (1): LIT_032

Cases (1): Totarabank Eco-Village