Sep 19, 2019 |
Read time: 4 mins
The quality of the wind resource assessment campaign is usually judged against its ability to obtain debt leverage into a project, i.e. the bankability of the resulting energy yield estimates. Usually, references are made to IEC standard 61400-12-1 and MEASNET guidelines for evaluation of site-specific wind conditions Ver. 2 April 2016. to assess if the wind resource assessment campaign has been designed and executed to provide sufficient high quality wind data. One such requirement specified by MEASNET is to have a minimum of 12 consecutive months of measurement period from any of the local meteorological mast to perform the long term assessment and energy yield calculation. A requirement directly reflected by major institutional debt providers, such as the European Investment Bank.
However, as a developer, blindly relying on 12 months as a gauge for an unbiased wind resource campaign could be risky. The reference year for the 12-month measurement period, could by itself introduce a long term bias on the predicted P50 production, if the specific 12-month period is observing “unusual” wind speeds, e.g. significantly higher or lower wind speeds on the long term average. Below, we demonstrate that point, based on a series of energy yield estimation conducted at a site with several years of local measurements.
To demonstrate how the concurrent measurement period influences the resulting long term P50 predictions, we analysed data from a site with four meteorological masts with data recorded from 2013-2018. Within the measurement period, we then identified the 12 consecutive months with having the highest and lowest average wind speed across the four met mast. These two 12-month measurement periods were then used as the input for the MCP process with long term data (ERA5) to conduct the energy yield estimates for the same wind farm.
The mean wind speed from the four masts shows the highest high wind for a consecutive 12 months period between January 2014 and December 2014 at 8.0m/s, and the lowest average wind speed from December 2014 and November 2015 at 7.4 m/s. The below graph depicts the data from the entire measurement campaign, and how the two 12-month periods are systematically above or below the average monthly wind speed reflected by the yellow line.
The collective one year lowest mean wind speed data and highest wind speed data are considered to evaluate the long term wind climate by correlating with the long term reference data which has 30 years of data availability.
After performing the long term correction and energy yield estimation for the two 12-month reference periods, the impact of the reference period on the long term yield predictions was evident.
- When using the high wind speed period (Jan 2014 – Dec 2014) for long term correction and energy estimation, the predicted wind speed for the low wind period (Dec 2014 – Nov 2015) has been overestimated by 1.1% when compared to the actual wind speed measured at the site.
- When using the low wind speed period (Dec 2014 – Nov 2015) for long term correction and energy estimation, the predicted wind speed for the high wind period (Jan 2014 – Dec 2014) has been underestimated by 1.97% when compared to the actual wind speed measured at the site.
More importantly, these overestimation or underestimations have a direct impact on the long term predicted AEP. When comparing the long term AEP (P50) predicted based on the high wind speed period and low wind speed period, it reveals a difference of 2.6% in annual yield.
|Description||Average mean wind speed [m/s]||AEP (P50)
|Long Term corrected wind speed based on low wind speed period measurement||8.4||853.6|
|Long Term corrected wind speed based on high wind speed period measurement||8.5||875.8|
Such difference resulting from a bias in the measurement period can have a significant impact on the long term viability of projects, as it influences not only the P50 but also the P90/P99 values. Thereby either under- or overestimating the amount of debt funding which can be obtained for a project or fixed hedge obligations. The first obviously undermines the long term viability of projects, while the latter results in lower RoE due to underutilized debt-leverage. If assuming the same uncertainties for both predictions, the bias from conducting the measurement campaign during the high wind period would result in the predicted P50 value turning into only having a 332% probability of exceedance if the actual wind pattern is more similar to the wind pattern observed during the low wind period. Thereby resulting in a systematic production gap year on year.
This points to some important observations for developer or investors evaluating evaluating the design of wind resource assessment campaigns and long term production estimates. First of all, a MEASNET compliant measurement campaign, with 12 consecutive months of local measurements, does not necessarily result in unbiased energy yield estimates.
Luckily, it is not a sophisticated step to understand if you are potentially exposed to a bias from the measurement period. A simple comparison with long term reference data, such as ERA5 or MERRA2, would provide a reference for understanding the wind pattern in the during your wind measurement campaign. A view on the long term reference would reveal that if the measurement campaign was conducted in 2015 (low wind speed period used above), it would be subject to a potential bias of underestimating production. While the measurement period in 2014, despite being the highest during the measurement campaign, is not significantly above the long term average.