Very Short Term Wind Speed Forecasting Using Multivariable Dense Data with WLS-MARMA Model

In this study, very short-term wind speed forecasting problem, which is quite important for the future's electricity market -wind forecasting control algorithms, is investigated. Recently, the multi-channel (spatial) methods which uses neighboring (from different locations) wind measurements are become popular. But it is not always possible to collect spatially distributed neighboring wind speed values around target location simultaneously. In this study, previously proposed multichannel autoregressive moving average (MARMA) model is applied to local multiple sensor measurements such as wind speed, direction, temperature, pressure, solar radiation etc. instead of neighboring (distributed) wind speed measurements. It is shown that weighted least squares solution based MARMA model (WLS-MARMA) can give more accurate wind speed estimation results according to other well-known benchmark methods (such as Persistence, AR, VAR) with real data set.