Battery Optimisation

Overview

The Battery Optimisation module provides a full operational dispatch optimisation engine for grid-scale Battery Energy Storage Systems (BESS). It combines real-time NEM price data, FCAS market signals, and ML price forecasts to produce an optimal 24-hour dispatch schedule that maximises revenue while respecting battery operating constraints.

Optimisation Problem

The battery optimisation is a Mixed Integer Linear Program (MILP) solved at each dispatch interval:

Maximise:
    Σ_t [price_t × (discharge_t - charge_t) × η_discharge
         + FCAS_rev_t × FCAS_enabled_t]
    - degradation_cost × (charge_t + discharge_t)

Subject to:
    SOC_t = SOC_(t-1) + charge_t × η_charge / capacity - discharge_t / (η_discharge × capacity)
    SOC_min ≤ SOC_t ≤ SOC_max
    0 ≤ charge_t ≤ max_charge_rate
    0 ≤ discharge_t ≤ max_discharge_rate
    charge_t × discharge_t = 0  (can't charge and discharge simultaneously)
    FCAS_enabled_t ≤ headroom_t  (FCAS limited by SOC headroom)

Input Data

Input	Source	Update Frequency
Current spot prices	gold.nem_prices_5min (Lakebase)	Every 5 min
Price forecast (30-min)	ML model via /api/forecasts	Every 30 min
FCAS prices	gold.nem_fcas_prices (Lakebase)	Every 5 min
Current SOC	DUID SCADA via API	Every 5 min
Pre-dispatch prices	AEMO pre-dispatch feed	Every 5 min

Dashboard Pages

Dispatch Strategy (/middle-office/battery-optimisation/strategy)

• 24-hour charge/discharge schedule (optimised)
• Current and projected SOC profile
• Revenue breakdown: energy arbitrage vs FCAS by hour
• Sensitivity analysis: how revenue changes with ±10% price assumptions

Screenshot: Battery optimisation dashboard showing 24h charge/discharge bars, SOC curve, and revenue attribution waterfall.

Live Arbitrage P&L (/middle-office/battery-optimisation/pnl)

Realised P&L tracked against the optimised schedule:

Actual_PnL = Σ_t (actual_dispatch_price_t × actual_output_mw_t × 5min/60)
Schedule_PnL = Σ_t (forecast_price_t × scheduled_mw_t × 5min/60)
Tracking_Error = Actual_PnL - Schedule_PnL

Tracking error arises from price forecast errors, rebids by other participants, and FCAS price movements.

SOC Management (/middle-office/battery-optimisation/soc)

• Real-time SOC indicator (0–100%) with min/max constraint lines
• Historical SOC profile — last 7 days
• Cycle depth histogram — distribution of discharge depths
• Degradation accumulation — estimated capacity loss to date

FCAS Co-optimisation

A key advantage of BESS in the NEM is simultaneous FCAS provision:

Simultaneous Services

A battery with 100 MW / 2-hour capacity can typically provide:

• Energy arbitrage: 60 MW discharge (peak prices) / 60 MW charge (off-peak)
• RAISEREG: 20 MW headroom reserved for frequency regulation raise
• LOWERREG: 20 MW headroom reserved for frequency regulation lower
• RAISE6SEC: 10 MW available for 6-second contingency response
• LOWER6SEC: 10 MW available for 6-second contingency response

The FCAS trapezium constraint means FCAS enablement reduces energy dispatch headroom — the optimiser balances this trade-off.

# FCAS trapezium feasibility check
def check_fcas_trapezium(
    dispatch_mw: float,
    raise_reg_mw: float,
    lower_reg_mw: float,
    max_capacity: float,
    min_capacity: float = 0
) -> bool:
    """
    Validates FCAS offer is feasible given current dispatch level.
    Returns True if the FCAS offer is within the trapezium constraints.
    """
    max_output = dispatch_mw + raise_reg_mw
    min_output = dispatch_mw - lower_reg_mw
    return (max_output <= max_capacity) and (min_output >= min_capacity)

Bid Generation

The optimisation output feeds directly into a bid file for AEMO submission (for operators using the API integration):

{
  "duid": "VBBG1",
  "trading_day": "2025-03-22",
  "bids": [
    {
      "period": "00:00",
      "energy_bands": {
        "band_1": {"price": -1000, "volume_mw": 100},
        "band_4": {"price": 200, "volume_mw": 100}
      },
      "fcas_bids": {
        "RAISEREG": {"availability_mw": 20, "enablement_min": 80, "enablement_max": 100},
        "LOWERREG": {"availability_mw": 20, "enablement_min": 0, "enablement_max": 20}
      }
    }
  ]
}

API Endpoints

# Generate optimised 24h dispatch schedule
POST /api/battery-opt/schedule
{
  "duid": "VBBG1",
  "capacity_mwh": 450,
  "power_mw": 300,
  "soc_current_pct": 55,
  "fcas_services": ["RAISEREG", "LOWERREG", "RAISE6SEC"],
  "price_forecast_source": "ml_model"  // or "pre_dispatch"
}

# Current SOC and status
GET /api/battery-opt/status?duid=VBBG1

# Realised vs scheduled P&L
GET /api/battery-opt/pnl?duid=VBBG1&date=2025-03-21

Tip

The battery optimisation engine uses the ML price forecast from the NEM Price Forecasting model (LightGBM, 30-min horizon) as the primary forecast input. For longer horizons (>6 hours), it falls back to ASX futures-derived forward prices.