New Forecasting Tools Enhance Wind Integration In Idaho And Oregon

Idaho Power Company’s new wind availability forecasting models are helping to integrate wind power more efficiently and cost effectively.
Published: Thu 25 Sep 2014

Over the past decade, Idaho Power Company (IPC) has experienced a large growth in wind power capacity, currently standing around 700MW and providing at any given time up to 35% of the company’s system needs.

Customized wind forecast models

With the high potential for wind integration and the absence of commercial products offering the needed level of accuracy, IPC opted to develop a Renewables Integration Tool (RIT) comprising a series of models and databases for forecasting weather conditions and the availability of wind energy resources. The project formed part of a larger smart grid project, supported with a Smart Grid Investment Grant.

The goal was to address three wind forecast intervals: within one hour, hourly, and day ahead. IPC investigated a number of approaches, eventually identifying a set of weather forecasting models and datasets developed by the University of Arizona that could be tailored for IPC’s specific needs.

RIT uses several weather models. The primary forecasting model for short-term forecasts runs four times a day and makes wind predictions 72 hours into the future. A second model, run once every day, Monday through Friday, makes predictions 180 hours into the future and is the primary forecast model for timeframes beyond 72 hours.

Weather data is drawn from meteorological towers located in multiple sites in five major wind parks, and from various public meteorological sites. The wind parks are geographically dispersed across about 500km (300 miles) of hilly terrain.

RIT also includes analysis and mapping tools that can graphically show wind speeds at 10m and 80m above the ground, abrupt changes in wind directions, and other important weather parameters.

Improved wind forecasting

Using RIT over the three months February-April 2014, the mean-absolute-hourly-error (MAE – the sum of absolute value of the observed amount of hourly wind generation, minus the forecasted amount, and divided by the total number of hours in the forecasting period) was 100MWh. This represents a 26% to 33% improvement in forecasting accuracy over IPC’s previous forecasting method (MAE 136MWh) and a rolling 3-day average of actual historical wind generation (MAE 149MWh).

However, IPC still sees a wide variability of energy received in any given hour due to the difficulty of predicting hourly wind characteristics. As a result, grid operators still confront significant levels of uncertainty due to unexpected variations in wind generation. While RIT creates higher confidence in wind generation forecasts, this confidence can vary from day-to-day as well as from hour-to hour.

Financial benefits of improved wind forecasting

Based on conservative figures of a 25% improvement in forecasting accuracy and an estimated US$15/MW, IPC estimates that using the RIT saved about US$287,000 for the three-month period – or about US$96,000 per month – as compared to the two other forecasting methods that were used before RIT was developed.

Over time, these savings outweigh the costs of developing, maintaining and upgrading the RIT.

Future developments

IPC is continuing to develop the RIT. More and better weather data is still needed to support advanced forecasting tools. IPC plans to expand data collection to more wind parks, and include weather and operational data requirements in new power purchase agreements for wind energy resources.

Managing the large volumes of weather, wind turbine performance, and other system-related data on electricity demand and power supplies requires continuing efforts to develop algorithms for processing and analysis. As a result, improvements in data analytics remain an important priority for the future.  

Although the RIT is a customized software platform, utilities interested in developing comparable wind forecasting capabilities could use it as a template. However, weather data, wind turbine performance information, and statistical algorithms would have to be created to suit local conditions.

Further reading

US Department of Energy: New Forecasting Tools Enhance Wind Energy Integration In Idaho And Oregon