Define the Goals of Your Generation Forecast System

The goals for your forecasting system can be expressed as a reduction of forecasting errors for different time horizons, based on a variety of metrics. However, it may also be valuable to express your goals in more immediate terms such as the quantity and types of system operating reserves needed at any given time. Which metrics you choose and how they are considered depends on the specific needs of your system and the amount of data available.

---

Main Points

• Renewable energy forecasting errors are typically evaluated using a standard set of performance metrics. Which metrics are used depends on the technology and the forecast time horizon, as well as factors specific to the improvement you want to achieve.
• Goals for improving an existing system can be defined in terms of percentage reduction in forecast errors compared to the previous system, although a variety of other goals can be set for new or improved systems, such as reduction in needed operating reserves or VRE curtailment.
• These goals will influence how various forecasting errors are weighted and considered.
• Setting realistic goals will also depend on the capability of your forecasting system and the amount and quality of data available.

---

First, Review This Example

Forecasting metrics can consist of various statistical tests or other evaluations assessed over different time horizons. Separate metrics apply for solar PV and wind generators. The metrics used can be customized to your particular needs and expanded as more data becomes available.

Next, Read More About Types of Metrics

Now, See This Table

Forecasting targets vary depending on the time horizon considered. The table below is an example showing baseline and target solar power forecasts for Day Ahead (DA), Four Hour Ahead (4HA), One Hour Ahead (1HA), and 15 Minute Ahead (15MA) forecast horizons using a variety of metrics. The definitions of the metrics used are in Table 1 on pages 3 – 4 of the referenced publication.

See Table from page 11 of Baseline and Target Values for Regional and Point PV Power Forecasts: Toward Improved Solar Forecasting.

Next, Read This

A performance scheme reflects the importance of forecast parameters that can be improved upon over time, with incentives and resources allocated to the improvement. Evaluation of forecast parameters should depend on:

1. the objective of the forecasting solution;
2. the use/application of the forecasts; and
3. the available input at the forecast generation time.

Finally, Read This

Weighting the different forecast parameters by which ones are most important for your system creates a cost function or evaluation matrix you can use to evaluate your forecasting system’s performance.

Suggested Actions & Next Steps

• Produce an initial list of achievable goals (e.g., average day-ahead solar generation forecast error of less than 2% Normalized RMSE).
• Discuss goals and metrics with the Technical Review Committee and revise them based on TRC feedback.
• Consider how you would weight your performance goals in an evaluation framework you will use to test the performance of your forecasting system.
• Continue to revise the goals and evaluation framework and inform the TRC as new information becomes available.