Probabilistic Forecasting of Hourly Wind Power Load in South Africa

Abstract

Forecasters must make some educated assumptions about future electricity demand and effectively explain them because real electricity demand and forecasted electricity demand always differ. In order to forecast future wind power load base stations in South Africa, the authors evaluate hourly wind power generation in this chapter. The predicted time series can be used to show the flow of the load demand trend for electricity. Because of the noise in the original time series, they provide the enhanced gradient-boosted decision tree algorithm based on Kalman filter (GBDT-KF). They compare how well the proposed GBDT-KF method performs with varying numbers of decision trees. Re-sampling is cross-validated three times at a 10 fold interval. The employment of MSE, RMSE, MAE, and MAPE allowed for the selection of the best model. As a result, the total findings showed that 90% of the test data and 92% of the training data were successful. The findings of this study will be helpful to the energy sector and decision-makers for planning and future use in the hourly electricity forecasting domain.