The authors used a three-year hourly wind speed dataset to test the model against SVR . The reported model had smaller MSE and MAPE than the SVR model. However, the user defined threshold ? and the offset ω were not clearly defined and it AG 556 is dataset dependent.
For solar irradiance forecasting, a multi-stage ANN was reported in . There were three stages: (i) first stage forecasts the average atmospheric pressure based on the historical average atmospheric pressure data, (ii) second stage forecasts the irradiance level based on the output of the first stage and (iii) third stage forecasts the solar irradiance based on the output of the second stage as well as the historical meteorological data. This multi-stage ANN was applied to forecast the heat pump water supply system rather than a PV system. The results showed a reasonable improvement compared with a single-stage ANN.
4. Results and discussion
In effector section, selected ensemble forecasting methods are evaluated with wind speed/power and solar irradiance time series data. Firstly the results reported from selected literature are summarized and compared and secondly three selected ensemble forecasting methods are evaluated with four wind speed time series datasets and four solar irradiance time series datasets and their performance is compared and discussed.