Demand & Weather

Overview

Weather is the primary driver of electricity demand in Australia — air conditioning load on hot summer days can push NEM demand to record highs, while mild autumn days produce demand troughs. Energy Copilot integrates Bureau of Meteorology (BOM) weather data for all NEM regions with demand data from AEMO to provide comprehensive demand-weather analytics.

Weather Data Sources

Energy Copilot uses the Open-Meteo BOM ACCESS-G weather model, which provides:

• Historical weather: observed temperature, humidity, wind speed, irradiance (1 year lookback)
• Short-term NWP forecasts: +72 hours at 1-hour resolution
• Medium-term forecasts: +7 days at 3-hour resolution

Weather stations are mapped to NEM regions using population-weighted averages:

NEM Region	Primary Weather Location	Population Centre
NSW1	Sydney (Observatory Hill)	Greater Sydney
QLD1	Brisbane (RAAF Amberley)	South East QLD
SA1	Adelaide (Kent Town)	Greater Adelaide
TAS1	Hobart (Ellerslie Road)	Greater Hobart
VIC1	Melbourne (Olympic Park)	Greater Melbourne

Dashboard Pages

Weather Overview (/front-office/weather)

• Current temperature map: colour-coded by temperature for all 5 regions
• Temperature forecast chart: next 7 days with min/max/mean
• BOM ACCESS-G model parameters: temperature, humidity, cloud cover, wind speed, solar irradiance

Temperature-Demand Correlation (/front-office/demand-weather-correlation)

• Scatter plot: temperature vs demand for each region (historical 90 days)
• R² coefficient displayed for each region’s temperature-demand relationship
• Breakpoint detection: temperature threshold where cooling load activates
• Heating vs cooling asymmetry: demand response is typically larger for cooling

Demand Forecasting (/front-office/demand-forecast)

The ML demand forecast uses LightGBM with weather features:

• Inputs: temperature forecast, day-of-week, public holiday flag, season, lag features (demand at t-1, t-7, t-48 periods)
• Output: 30-minute demand forecast for each region, 24-hour horizon
• MAPE target: <3% for operational planning horizon

Screenshot: Demand forecast chart for NSW1 showing actual demand (solid) vs ML forecast (dashed) with 90% prediction interval band.

Cooling and Heating Degree Days

Cooling Degree Days (CDD) and Heating Degree Days (HDD) are standard energy sector metrics:

CDD = max(0, Daily_Mean_Temp - 18°C)
HDD = max(0, 18°C - Daily_Mean_Temp)

Energy Copilot calculates and displays:

• Weekly and monthly CDD/HDD by region
• Cumulative CDD/HDD for the current summer/winter season
• Comparison to 5-year historical average

Demand Records and Extremes

The platform tracks demand records:

Region	Summer Peak Record	Date	MW
NSW1	All-time peak	Jan 2011	14,846
QLD1	All-time peak	Feb 2004	9,673
SA1	All-time peak	Jan 2009	3,442
VIC1	All-time peak	Jan 2009	10,854
NEM Total	All-time peak	Jan 2011	~35,000

Note

These historical peak records pre-date significant rooftop solar uptake. Operational demand (excluding rooftop solar) has been declining even as population and economic activity grow.

Gold Table Schema

-- gold.weather_nem_regions
SELECT
    observation_datetime,  -- TIMESTAMP: UTC observation time
    interval_date,         -- DATE: partition key
    region_id,             -- STRING: NEM region
    temperature_c,         -- DOUBLE: temperature (°C)
    humidity_pct,          -- DOUBLE: relative humidity (%)
    wind_speed_ms,         -- DOUBLE: wind speed (m/s)
    wind_direction_deg,    -- DOUBLE: wind direction (degrees)
    solar_irradiance_wm2,  -- DOUBLE: global horizontal irradiance (W/m²)
    precipitation_mm,      -- DOUBLE: precipitation (mm)
    is_forecast,           -- BOOLEAN: true for NWP forecast periods
    forecast_horizon_h     -- INT: hours ahead (NULL for observed)
FROM energy_copilot.gold.weather_nem_regions
WHERE interval_date = CURRENT_DATE()
ORDER BY observation_datetime DESC;

API Endpoints

# Current weather for all regions
GET /api/weather/latest

# Weather history for specific region
GET /api/weather/history?region=SA1&start=2025-03-01

# Weather forecast (NWP)
GET /api/weather/forecast?region=NSW1&horizon=72h

# Demand forecast
GET /api/demand/forecast?region=VIC1&horizon=24h

# Demand-weather correlation stats
GET /api/demand/weather-correlation?region=QLD1&period=90d

Tip

Ask the copilot: “Is there a heatwave forecast for SA this week and what is the expected demand impact?” The copilot will retrieve the weather forecast and apply demand-temperature correlation models to estimate peak demand and price risk.