Short-term wind power forecasting based on numerical weather prediction adjustment

Most wind power forecasting methods today take numerical weather prediction (NWP) as their inputs. Therefore, the accuracy of these forecasting methods highly depends on the accuracy of NWP. This paper involves in studying the statistical features of NWP. A total of four error patterns are pre-defined according to the statistical features of NWP. Moreover, an advanced autoregressive integrated moving average (ARIMA) simulator with error information integrated is established to adjust the NWP. Finally, a pair of comparison tests based on support vector machine (SVM) is run with raw NWP and adjusted NWP as inputs respectively. It proves that the adjusted NWP increases forecast accuracy greatly.