An Open Model for Generating High Resolution Wind Power Production Scenarios

Abstract

Efficient integration of variable renewable energy (VRE) such as wind power into power systems requires methods for power system operation planning that account for VRE uncertainty and variability. This has motivated extensive research into unit commitment (UC) and optimal power flow (OPF) formulations with VRE uncertainty. However, these formulations are often tested using significantly simplified representations of VRE production. We seek to address this issue by providing a model for generating realistic wind power scenarios using real production and forecast data. The scenarios are generated using 5-min production and 30-min forecast data for real wind farms from Australia. The model captures the empirical distribution of the forecast errors and the covariance between different wind farms. The high time resolution of the production data also allows the recreation of the high-frequency (5-min) component of wind power production. The resulting model is openly available, and can be used to generate wind power scenarios for use in formulations for operation planning of power systems (UC/OPF) considering wind uncertainty and intra-hour variability. The scenarios can be tailored according to preferences for, e.g., the number of wind farms and their geographical dispersion.

Capacity and Intra-Hour Ramp Reserves for Wind Integration

“…6 shows the normalized demand profile, which was obtained from CAISO. The wind power production scenarios were generated using the model described in [38], and Fig. 7 shows the wind power scenarios used in the UC formulations.…”

“…7 shows the wind power scenarios used in the UC formulations. The wind scenarios have 5-min time resolution and are based on the empirical distribution of forecast errors, while preserving the correlation of the forecast errors between different time periods and different wind farms [38].…”

“…Also, we test formulations with a combination of hourly and intra-hourly ramp reserves as well as formulations with only intra-hourly ramp reserves. The formulations are evaluated using wind power production scenarios from the model presented in [38]. The scenarios have 5-min time resolution and are based on data from real wind farms, thus capturing the empirical distribution of forecast errors and correlation between wind farms.…”