# Battery Duration Comparison: 2-Hour vs 4-Hour in Irish SEM

**Status:** Complete | **Confidence:** HIGH for methodology, HIGH for conclusion
**Data:** 64,670 hourly SEM prices (Oct 2018 – Feb 2026), 2,695 backtest days
**Scripts:** `scripts/backtest_duration_comparison.py`, `scripts/analyze_duration.py`
**Output:** `data/backtest_2h_results.csv`, `data/backtest_4h_results.csv`, `data/duration_comparison.json`

**Bottom line: 4-hour wins.** The 4h battery earns ~61% more revenue per MW than 2h, while costing only ~33% more. The 4h system has a higher ROI in every year, every strategy, and every scenario tested — including when CRM capacity payments are stacked. The Irish market has wide enough daily price troughs and peaks (8-9 cheap hours, 7 expensive hours) that hours 3 and 4 still capture ~85% of the per-hour spread of hours 1 and 2.

---

## Section 1: Why Duration Matters

Duration is the single most consequential design parameter for a battery project. It determines:

1. **Revenue ceiling.** A 2h battery at 50 MW can discharge 100 MWh per cycle; a 4h battery discharges 200 MWh. If spreads are wide enough to fill both, the 4h earns roughly twice the gross revenue per cycle.

2. **CAPEX.** Battery cells scale linearly with MWh, but the power conversion system (PCS), grid connection, transformer, EPC, and development costs scale primarily with MW. Doubling duration does NOT double cost — it adds ~33% for cell-related components while fixed MW-linked costs stay constant.

3. **Cycling behaviour.** Shorter batteries can cycle more frequently (a 2h battery needs only 4 hours for a full charge-discharge cycle vs 8 hours for 4h). But each cycle captures a smaller spread.

4. **Capacity market payments.** De-rating factors increase with duration: ~25% for 2h vs ~40% for 4h in the SEM CRM. This gives 4h batteries 60% more capacity revenue per MW installed.

The question is whether the extra revenue from hours 3-4 of storage justifies the extra cell cost. This analysis answers that question using 7+ years of actual SEM price data.

---

## Section 2: Backtest Configuration

Both configurations tested:

| Parameter | 2-Hour | 4-Hour |
|-----------|--------|--------|
| Power rating | 50 MW | 50 MW |
| Energy capacity | 100 MWh | 200 MWh |
| Usable capacity (5-95% SoC) | 90 MWh | 180 MWh |
| Round-trip efficiency | 85% AC-AC | 85% AC-AC |
| Fixed schedule charge | 03:00-04:59 | 02:00-05:59 |
| Fixed schedule discharge | 17:00-18:59 | 17:00-20:59 |
| LP strategy | Perfect foresight, daily | Perfect foresight, daily |

Methodology is identical to `BACKTEST-RESULTS.md`. SoC resets daily. RTE applied on charge. Two strategies: perfect foresight (PF, LP upper bound) and fixed schedule (FS, zero-intelligence lower bound).

---

## Section 3: Year-by-Year Revenue Comparison

### Revenue per MW (EUR/MW/year) — arbitrage only

| Year | 2h PF | 4h PF | 2h/4h | 2h FS | 4h FS | 2h/4h |
|------|------:|------:|------:|------:|------:|------:|
| 2019 | 33,089 | 53,045 | 62.4% | 21,743 | 35,098 | 61.9% |
| 2020 | 32,120 | 50,537 | 63.6% | 23,126 | 36,784 | 62.9% |
| 2021 | 68,635 | 110,403 | 62.2% | 50,572 | 76,992 | 65.7% |
| 2022 | 86,903 | 141,576 | 61.4% | 52,993 | 86,150 | 61.5% |
| 2023 | 52,082 | 83,564 | 62.3% | 30,276 | 48,987 | 61.8% |
| 2024 | 60,788 | 99,810 | 60.9% | 34,762 | 58,202 | 59.7% |
| 2025 | 69,400 | 115,308 | 60.2% | 35,995 | 61,857 | 58.2% |
| **Avg (all)** | **56,619** | **91,246** | **62.1%** | **36,139** | **57,602** | **62.7%** |
| **Avg (ex-crisis)** | **49,569** | **79,665** | **62.2%** | **30,924** | **49,612** | **62.3%** |

The 2h battery consistently earns **~62% of 4h revenue per MW**. This ratio is remarkably stable across years and strategies — it barely changes between the crisis years (2021-22) and normal years.

### Revenue per MWh of storage (EUR/MWh-installed/year)

| Year | 2h PF | 4h PF | 2h/4h | 2h FS | 4h FS | 2h/4h |
|------|------:|------:|------:|------:|------:|------:|
| 2019 | 16,545 | 13,261 | 124.8% | 10,871 | 8,774 | 123.9% |
| 2020 | 16,060 | 12,634 | 127.1% | 11,563 | 9,196 | 125.7% |
| 2021 | 34,317 | 27,601 | 124.3% | 25,286 | 19,248 | 131.4% |
| 2022 | 43,452 | 35,394 | 122.8% | 26,497 | 21,537 | 123.0% |
| 2023 | 26,041 | 20,891 | 124.7% | 15,138 | 12,247 | 123.6% |
| 2024 | 30,394 | 24,952 | 121.8% | 17,381 | 14,551 | 119.5% |
| 2025 | 34,700 | 28,827 | 120.4% | 17,997 | 15,464 | 116.4% |

The 2h battery earns **~120-125% per MWh of installed storage** compared to 4h. This means each MWh of the 2h system works harder — but not hard enough to overcome the cost advantage of the 4h system (see Section 4).

### Cycles per day

| Year | 2h PF | 4h PF | 2h FS | 4h FS |
|------|------:|------:|------:|------:|
| 2019 | 1.78 | 1.45 | 0.94 | 0.94 |
| 2020 | 1.98 | 1.49 | 0.94 | 0.94 |
| 2021 | 1.63 | 1.35 | 0.94 | 0.94 |
| 2022 | 1.64 | 1.34 | 0.94 | 0.94 |
| 2023 | 1.74 | 1.45 | 0.94 | 0.94 |
| 2024 | 1.72 | 1.48 | 0.94 | 0.94 |
| 2025 | 1.73 | 1.52 | 0.94 | 0.94 |
| **Average** | **1.72** | **1.42** | **0.94** | **0.94** |

The 2h battery cycles ~21% more frequently (1.72 vs 1.42 cycles/day with perfect foresight). This is the mechanism by which it partially compensates for lower energy capacity — it runs more cycles per day, each of shorter duration. However, marginal cycles capture diminishing spreads, so 21% more cycles does not translate into 21% more revenue per MWh.

The fixed schedule achieves 0.94 cycles/day for both durations — one cycle per day is the same regardless of duration (the limiting factor is the fixed window, not the battery size).

---

## Section 4: Revenue per EUR of CAPEX — The Real Question

### CAPEX build-up

From `CAPEX-BUILD-UP.md`, decomposing the EUR 34.0M reference case (50 MW / 200 MWh):

**MWh-linked costs (scale with duration):**

| Component | 200 MWh (EUR M) | 100 MWh (EUR M) |
|-----------|----------------:|----------------:|
| LFP cells | 6.8 | 3.4 |
| Pack assembly | 2.0 | 1.0 |
| Containers, HVAC, fire suppression | 3.4 | 1.7 |
| BMS | 1.7 | 0.85 |
| PCS (MWh-linked portion) | 0.55 | 0.28 |
| Shipping + insurance + duty | 0.8 | 0.4 |
| **Subtotal MWh-linked** | **15.25** | **7.63** |

**MW-linked costs (same for both durations):**

| Component | EUR M |
|-----------|------:|
| PCS (MW-linked portion) | 0.55 |
| Grid connection | 4.50 |
| Transformer | 2.80 |
| EPC / installation | 5.10 |
| Development costs | 0.80 |
| Legal & permitting | 0.30 |
| Insurance (construction) | 0.30 |
| MEC bond | 1.25 |
| **Subtotal MW-linked** | **15.60** |

**Totals:**

| Metric | 2-Hour | 4-Hour |
|--------|-------:|-------:|
| MWh-linked | 7.6 | 15.3 |
| MW-linked | 15.6 | 15.6 |
| Subtotal | 23.2 | 30.9 |
| Contingency (10%) | 2.3 | 3.1 |
| **Total CAPEX** | **EUR 25.5M** | **EUR 33.9M** |
| Per MW installed | EUR 511k/MW | EUR 679k/MW |
| Per kWh installed | EUR 256/kWh | EUR 170/kWh |
| **Cost saving for 2h** | **EUR 8.4M (24.7%)** | — |

The 2h battery saves EUR 8.4M (25%) — almost entirely from halving the cell count. All MW-linked infrastructure (grid, transformer, EPC, development) is identical.

### ROI comparison (arbitrage only)

| Scenario | 2h ROI | 4h ROI | Winner |
|----------|-------:|-------:|--------|
| PF (all years) | 11.1% | 13.5% | **4h** |
| PF (ex-crisis) | 9.7% | 11.7% | **4h** |
| FS (all years) | 7.1% | 8.5% | **4h** |
| FS (ex-crisis) | 6.1% | 7.3% | **4h** |

| Scenario | 2h Payback | 4h Payback | Winner |
|----------|----------:|----------:|--------|
| PF (all years) | 9.0 yrs | 7.4 yrs | **4h** |
| PF (ex-crisis) | 10.3 yrs | 8.5 yrs | **4h** |
| FS (all years) | 14.1 yrs | 11.8 yrs | **4h** |
| FS (ex-crisis) | 16.5 yrs | 13.7 yrs | **4h** |

**The 4h battery wins on ROI in every scenario.** The 2h saves 25% on CAPEX but loses 38% of revenue — the revenue loss exceeds the cost saving.

### Why 2h loses despite lower cost

The arithmetic is simple:
- **2h earns 62% of 4h revenue** (EUR 56,619 vs EUR 91,246 per MW, PF average)
- **2h costs 75% of 4h CAPEX** (EUR 25.5M vs EUR 33.9M)
- Revenue ratio (62%) < Cost ratio (75%) → 4h has better ROI

For 2h to win, it would need to earn >75% of 4h revenue per MW. It earns only 62%. The gap is ~13 percentage points, which is too large to overcome.

---

## Section 5: CRM Capacity Payments Widen the Gap

The SEM Capacity Remuneration Mechanism (CRM) de-rates batteries by duration. Longer batteries get more capacity credit because they can sustain output for longer during system stress events.

| Duration | De-rating factor | CRM revenue (at EUR 149,960/MW de-rated) |
|----------|----------------:|------------------------------------------:|
| 2-hour | 25% | EUR 37,490/MW/yr |
| 4-hour | 40% | EUR 59,984/MW/yr |
| **4h advantage** | | **EUR 22,494/MW/yr** |

### Stacked ROI (arbitrage + CRM)

| Scenario | 2h Revenue | 4h Revenue | 2h ROI | 4h ROI | Winner |
|----------|----------:|----------:|-------:|-------:|--------|
| PF arb (avg) + CRM | 94,109 | 151,230 | 18.5% | 22.3% | **4h** |
| PF arb (ex-crisis) + CRM | 87,059 | 139,649 | 17.1% | 20.6% | **4h** |
| FS arb (avg) + CRM | 73,629 | 117,586 | 14.4% | 17.3% | **4h** |
| FS arb (ex-crisis) + CRM | 68,414 | 109,596 | 13.4% | 16.2% | **4h** |

CRM payments widen the gap. The 4h battery earns EUR 22.5k/MW/yr more from CRM than the 2h — this is roughly the same magnitude as the CAPEX savings of 2h (EUR 8.4M / 15 years = EUR 11.2k/MW/yr amortised). In other words, the CRM advantage of 4h roughly pays for the extra cells within the battery lifetime.

---

## Section 6: Irish Market Features — How Wide Are Peaks and Troughs?

The fundamental question for duration selection: are there 4+ hours of low prices and 4+ hours of high prices each day? Or is value concentrated in 1-2 hours?

### Daily price shape

Analysing all 2,695 days in the dataset, counting hours significantly below and above the daily mean:

| Year | Avg cheap hours | Avg expensive hours |
|------|----------------:|--------------------:|
| 2019 | 8.3 | 7.1 |
| 2020 | 8.0 | 7.3 |
| 2021 | 8.5 | 7.2 |
| 2022 | 8.2 | 7.6 |
| 2023 | 8.7 | 7.3 |
| 2024 | 9.0 | 7.3 |
| 2025 | 9.1 | 7.1 |

The Irish SEM has **8-9 cheap hours and 7 expensive hours per day on average**. This is wide enough to easily fill a 4-hour battery in both charge and discharge windows. The trend is toward *wider* troughs (more cheap hours), likely due to increasing wind penetration depressing overnight/shoulder prices.

### Spread concentration by duration

The "best N hours" spread measures the average price difference between the N most expensive hours and N cheapest hours of each day:

| Year | Best 1h | Best 2h | Best 3h | Best 4h | 2h/4h spread |
|------|--------:|--------:|--------:|--------:|--------------:|
| 2019 | 55.7 | 51.4 | 47.2 | 43.6 | 117.8% |
| 2020 | 53.4 | 48.1 | 43.7 | 40.1 | 119.8% |
| 2021 | 122.9 | 113.0 | 104.3 | 96.5 | 117.1% |
| 2022 | 165.6 | 154.3 | 144.5 | 135.4 | 114.0% |
| 2023 | 93.0 | 86.8 | 81.3 | 76.2 | 113.9% |
| 2024 | 101.6 | 95.8 | 89.7 | 84.1 | 114.0% |
| 2025 | 111.1 | 105.2 | 99.3 | 93.3 | 112.8% |

The 2h spread is only **13-18% wider** than the 4h spread. This means the 3rd and 4th cheapest/most expensive hours have spreads that are ~85% as large as the 1st and 2nd. Hours 3-4 are **not marginal** in the Irish SEM — they still capture substantial value.

### Marginal value of hours 3 and 4

| Year | 2h average spread | 4h average spread | Marginal spread (hrs 3-4) | Marginal as % of 2h spread |
|------|------------------:|------------------:|--------------------------:|---------------------------:|
| 2019 | 51.4 | 43.6 | -7.8 | -15.1% |
| 2020 | 48.1 | 40.1 | -7.9 | -16.5% |
| 2021 | 113.0 | 96.5 | -16.5 | -14.6% |
| 2022 | 154.3 | 135.4 | -18.9 | -12.2% |
| 2023 | 86.8 | 76.2 | -10.6 | -12.2% |
| 2024 | 95.8 | 84.1 | -11.7 | -12.2% |
| 2025 | 105.2 | 93.3 | -11.9 | -11.3% |

The average spread drops by only ~12-15% when extending from 2h to 4h. The 4h spread per hour (EUR 93/4 = EUR 23.3/MWh in 2025) is still excellent. Hours 3 and 4 are not "scraping the bottom" — they are earning at ~85% of the per-hour rate of hours 1 and 2.

**This is the key market feature that makes 4h viable in Ireland.** The daily price shape has broad peaks and troughs, not sharp spikes. Wind-driven low prices persist for many hours (typically 6-10 hours overnight), and evening peaks span 3-5 hours. A 4-hour battery can fill and empty without hitting the steep part of the price curve.

---

## Section 7: GB Precedent — What Duration Do Developers Choose?

### Current GB fleet

From `GB-BESS-PERFORMANCE.md`:

- **1-hour:** The dominant duration in GB's early fleet (2016-2023). Optimised for frequency response (DC, DM, DR) which required availability rather than energy throughput.
- **2-hour:** Becoming the new standard for merchant projects. Modo Energy's benchmark index is based on 2-hour assets. Cycling rate: 1.0 cycles/day average.
- **4-hour:** Very limited operational data in GB. Few assets deployed because frequency response (the dominant revenue stream until 2023) did not reward duration.

### Why GB chose short duration

GB batteries were historically optimised for **frequency response markets**, which pay for availability (MW), not energy throughput (MWh). A 1-hour battery earns the same frequency response revenue per MW as a 4-hour battery, at a fraction of the cost. As frequency response revenues collapsed (2022-2024, falling ~80%), GB operators shifted toward wholesale arbitrage, where duration matters more.

Cornwall Insight (2024) reports 2-hour assets earning approximately GBP 96k/MW/yr including capacity payments — suggesting the market is already rewarding longer duration.

### Why Ireland may favour longer duration

Ireland differs from GB in several ways that favour 4-hour:

1. **CRM de-rating rewards duration.** The SEM capacity mechanism explicitly de-rates shorter batteries. A 4h battery gets ~60% more capacity revenue per MW installed (EUR 60k vs EUR 37.5k at current clearing prices).

2. **DS3/DASSA is transitioning.** The current DS3 regime (ending Sep 2027) pays for availability regardless of duration. The replacement (DASSA) is expected to include energy-limited participation, where longer duration provides more flexibility.

3. **Wider daily price spreads.** Irish SEM spreads are larger in absolute terms than GB (EUR 93/MWh for best-4h spread vs ~GBP 54/MWh in GB). The extra hours of storage capture meaningful additional revenue.

4. **Higher wind penetration.** Ireland's ~42% RES-E (vs ~35% GB) creates longer periods of low prices when wind is high and longer periods of high prices when wind drops. Extended low-price windows mean a 4-hour battery can charge at low prices for the full 4 hours, not just 2.

5. **Smaller market, higher saturation risk.** With only ~6 GW peak demand, Ireland's ancillary services markets will saturate faster. Pure-availability revenue (which favours short duration) will compress sooner, pushing the revenue stack toward arbitrage (which favours longer duration).

### Emerging GB trend

The GB market is itself moving toward longer duration:
- Modo Energy notes a second daily cycle for 2hr batteries increases revenue by ~42%
- 1hr batteries are cycling 1.2 times/day, 2hr at 1.0 — suggesting the market rewards longer cycles
- New GB projects in planning increasingly target 2-hour duration
- Some developers exploring 4-hour for capacity market de-rating benefits

---

## Section 8: Recommendation

### Verdict: **4-hour duration is the correct design choice for the Irish SEM.**

**Confidence: HIGH**

### The numbers

| Metric | 2h (50 MW / 100 MWh) | 4h (50 MW / 200 MWh) |
|--------|----------------------:|----------------------:|
| CAPEX | EUR 25.5M | EUR 33.9M |
| CAPEX per MW | EUR 511k | EUR 679k |
| Arbitrage revenue (PF avg) | EUR 56,619/MW/yr | EUR 91,246/MW/yr |
| Arbitrage revenue (FS ex-crisis) | EUR 30,924/MW/yr | EUR 49,612/MW/yr |
| CRM revenue | EUR 37,490/MW/yr | EUR 59,984/MW/yr |
| Total revenue (PF avg + CRM) | EUR 94,109/MW/yr | EUR 151,230/MW/yr |
| Total revenue (FS ex-crisis + CRM) | EUR 68,414/MW/yr | EUR 109,596/MW/yr |
| Simple payback (PF avg, arb only) | 9.0 years | 7.4 years |
| Simple payback (FS ex-crisis, arb only) | 16.5 years | 13.7 years |
| Annual ROI (PF avg + CRM) | 18.5% | 22.3% |
| Annual ROI (FS ex-crisis + CRM) | 13.4% | 16.2% |

### Why the answer is unambiguous

For 2h to beat 4h on ROI, one of these conditions would need to hold:

1. **Revenue ratio > Cost ratio.** The 2h would need to earn >75% of 4h revenue per MW. It earns 62%. The gap is 13 percentage points — too large for any plausible market shift to close.

2. **CRM de-rating equal.** If both durations got the same de-rating, the CRM advantage of 4h (EUR 22.5k/MW/yr) would disappear. This is not how the SEM CRM works.

3. **Dramatically higher cycling for 2h.** The 2h PF already achieves 1.72 cycles/day vs 1.42 for 4h — a 21% advantage. To close the ROI gap, the 2h would need ~2.5+ cycles/day, which is physically possible but would require consistent multi-cycle opportunities that the data doesn't show.

4. **Cell costs much higher.** If cells were EUR 100+/kWh instead of EUR 34/kWh, the MWh-linked cost saving of 2h would be proportionally larger. At current cell prices (~EUR 34/kWh), halving cells saves EUR 7.6M — not enough to offset the EUR 17.3M revenue loss (annual, over 15 years).

None of these conditions hold or are likely to hold.

### One scenario where 2h could win

If CRM de-rating for 4h batteries falls (e.g., if the SEM methodology changes to penalise longer batteries), AND cell prices triple, AND the Irish price shape narrows to concentrated 1-2 hour peaks — then 2h might become competitive. This would require three simultaneous market/regulatory shifts, which is unlikely.

### Sensitivity: What if the price shape narrows?

The trend is actually in the opposite direction. The number of "cheap hours" per day has been increasing (from 8.0 in 2020 to 9.1 in 2025), consistent with rising wind penetration creating longer low-price windows. This makes 4h more attractive over time, not less.

---

## Appendix: Capture Rate Comparison

| Year | 2h Capture (FS/PF) | 4h Capture (FS/PF) |
|------|--------------------:|--------------------:|
| 2019 | 65.7% | 66.2% |
| 2020 | 72.0% | 72.8% |
| 2021 | 73.7% | 69.7% |
| 2022 | 61.0% | 60.9% |
| 2023 | 58.1% | 58.6% |
| 2024 | 57.2% | 58.3% |
| 2025 | 51.9% | 53.6% |

Capture rates are similar for both durations (~60-62% average). Neither duration has a systematic advantage in terms of how well the fixed schedule approximates perfect foresight. Both suffer the same declining trend as renewable penetration makes peak/trough timing less predictable.

---

*Analysis prepared 20 February 2026. Based on 2,695 days of SEM hourly price data. CAPEX estimates from CAPEX-BUILD-UP.md. CRM de-rating factors from SEM Committee methodology. GB operational data from GB-BESS-PERFORMANCE.md.*