2015 Volume 8 Issue 1 Pages 61-66
In wind-based energy operations, turbine performance, part deterioration rates and regular maintenance costs play a significant role in the economic viability of such projects. Problems to determine in advance the optimal future state of variables such as generator torque or blade pitch angle may be framed as anticipatory control problems, though successful application in real-world operations requires reliable forecasts of the future state of local wind speed. The authors propose a general probabilistic forecasting approach within the geographically robust CRPS minimization estimation framework suited to make use of AMeDAS weather variable observations, with the chief goal of evaluating the utility of the AMeDAS observations in the context of a rotor speed control problem where one seeks to maximize long-term power output at a specific turbine. Tested over a full year's worth of observations from sites across Japan, and using deterministic and non-deterministic evaluation metrics compared against standard references, the capacity of the proposed model as a forecaster at the 10-minute horizon was verified, and in doing so confirmed the potential wide-scale utility of the AMeDAS network in this and related anticipatory control problems.