P1.6
The impact of weather sensitivity on the economic value of ensemble forecasts

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 30 January 2006
The impact of weather sensitivity on the economic value of ensemble forecasts
Exhibit Hall A2 (Georgia World Congress Center)
Jing Yuan, Victoria University of Wellington, Wellington, New Zealand; and T. Simmers and J. McGregor

Poster PDF (526.5 kB)

The economic value of ensemble forecasts depends on the quality of the forecasts and the sensitivity of the business activity to the weather conditions. Good probabilistic forecasts may lead to better decisions and increase economic returns. However a forecast system, which generally scores well in traditional verifications, may not necessarily lead to better economic returns over time. In some cases, even a small number of inaccurate forecasts may lead to poor economic returns.

Using data from the National Centers for Environmental Prediction (NCEP) ensemble forecast system, the value of ensemble-based probabilistic forecasts with an idealized model of the electricity market is assessed by the method of utility maximization. Several forecast methods were used in a simple temperature driven model of electricity demand at cities around New Zealand and Australia. The results indicate that ensemble-based probabilistic forecasts are more reliable in forecasting electricity demand than the forecasts based on a single model. They can be used to help with decision making, reducing the overall risk. Bias correction, while helpful to traditional verification of weather forecasts, does not guarantee a significant improvement in economic returns if the profits are non-linearly related to the weather variables. In the model we considered, the penalty for over-estimating the temperature is greater than for under-estimating. The skill of business demand forecasts is sensitive to the ensemble spread, highlighting the importance of correcting for insufficient spread. Calibration of ensemble-based probabilistic forecasts is a significant issue and needs to consider both correction of ensemble mean and adjustment of ensemble spread. The calibrated ensemble-based probabilistic forecasts provide the best performance of all forecasts in our idealized model of the electricity market.