Regressing against local weather to improve PREDICTIONS

The GasTrac Feature

  • Local weather is recorded at the same time as appliance demand

What the Chart Shows

  • The two charts on the left side show gas demand from the furnace + hot water tank.
  • The two charts on the right show demand only from the furnace.
  • The charts on the top row (red) show appliance demand correlated to the regional airport weather. 
  • The charts on the bottom row (green) show appliance demand correlated to GasTrac ambient weather data.
  • Typically analysts will use all available data and correlate it to regional weather. As shown in the chart in the upper left, the correlation result is only 0.38. This is often considered too low to be usable in forecasting.
  • Removing the hot water demand results in a slight improvement (in the upper right chart).
  • Using GasTrac weather dramatically improves the correlation with demand, as shown in the two charts on the lower row. Correlation scores are now greater than 72% and this data is perfectly suited for developing forecasts.

Why it is Important

  • More than ever before accurate demand forecasts are important. Forecast accuracy can be improved significantly when local weather is used to correlate historic consumption.
  • The use of regional weather can result in significant forecast inaccuracy.

Key Point

  • Recording actual weather at the premise leads to improved correlations and forecast.