A data-mining based methodology for win forecasting

Short-term wind speed forecasting using Support Vector Machines

“…Higher or lower values lead to worst forecasting results. Table I presents the comparison between the forecasting results of the proposed SVM approach, including the best found parameterization for the considered case with both kernels, and several methodologies based on ANN which have been presented and compared in an authors' previous work [3]. Simulations 1 to 7 are based on the ANN approach, and simulations 8 to 15 are performed using the proposed SVM approach.…”

“…Figure 5 shows that the execution time of the SVM approach (regardless of the kernel that is used) increases when the amount of training data is larger. However, even the larger amount of training data (50 records, when the best parameterization of the eRBF kernel takes only 31 records, and the best parameterization of the RBF kernel takes only 4 records) leads to an execution time of nearly 23 seconds, less than half the time of the faster ANN approach, which takes about 50 seconds [3]. The SVM methodology using both of the kernel functions has been submitted to an exhaustive sensitivity analysis, from which the best parameterizations for the wind speed forecasting problem have been found.…”

“…Different kernels and the respective parameterizations are experimented, in order to achieve conclusions on which are the most advantageous combinations of parameters in order to attain the best results for the short-term wind speed forecasting problem. The results of the proposed approach are compared with those resulting from a previous approach, in which several models based on ANN are used and the comparison of the methods results is made using a real data base with 3 years historical data [3]. Results are compared in terms of the quality of the results (forecasting error) and also in what concerns the execution time of the forecasting models.…”

An integrated approach for distributed energy resource short-term scheduling in smart grids considering realistic power system simulation

“…An important input to the hour ahead problem is the energy resources status and the consumption and generation very-short term forecast. In the present case study, the information about energy resources status are sent by PSCAD ® and the forecast of consumption and generation is determined by a specific algorithm presented in [6]. Only with this information and with the information about the day ahead scheduling is possible to do the hour ahead scheduling to period t.…”

“…This is mainly due to the lack of accuracy in wind forecasting when the forecasting anticipation is increased. In [6] the authors demonstrate that wind forecasting can be very accurate for very short-term forecasting, using the last 5 h of wind speed data to predict the next 5 min. This methodology can be used in this case to update 5 min ahead optimization input data.…”