This Lessons Learnt Report outlines the key learnings Windlab Limited has learnt from the period between October 2019 – April 2020.
Report extract
This project proposes to understand a number of aspects of a system of upstream measurements and forecasting methodologies that would make use of information about the future wind that is embedded in the flow upstream of a wind farm. Some of the questions asked are technical or scientific in nature while others are more about the economic viability of implementing such a system. During this reporting period, two LIDARs installed at the Kiata Wind Farm continued to build a database of wind measurements from near the surface up to turbine hub height, at a number of locations around the wind farm. Using this data analysis was done and a number of experiments were undertaken to begin to answer the key questions posed by the project.
During this reporting period we found that wind trajectory tracing methods based on backtracking air parcel trajectories from wind turbine nacelles can be used to accurately produce a trajectory probability density map. This map can be used to locate the optimum locations for upstream measurements. In addition we found from simple one- turbine/one-direction machine learning forecast models, that low level measurements, as low as 10-20m, are effective in developing such a forecast system. This means that upstream measurements can be made quite inexpensively. This has significant implications for the economics of such a system. And finally, like all exploratory endeavours, we learned that the raw accuracy metrics from forecast algorithms were being strongly degraded by edge cases and incompatibilities with AEMO’s error calculation methodology – a situation we are currently addressing.