Prediction and aggregation of regional PV and wind generation based on neural computation and real measurements

Abstract

With the increasing integration of renewable energy sources (RES) in electricity grid networks, the power supply system has been changed to be decentralised. The complexity of the grids increases. So the estimation and prediction of regional renewable power integration is essential for grid stability and planning. However, the power prediction from single virtual power plant (VPP) is no longer suitable for the large scale regional renewable power generation. This paper proposed a prediction model based on neural computation. Various architectures of artificial neural network (ANN) are studied. Different input features are investigated and uncertainty of the prediction results are assessed. The relative mean absolute error (rMAE) to installed capacity can reach 2.5% in summer and 0.5% in winter for PV power prediction. By using the reduced mean values of input categories features the rMAE can reach 1.7% and 4.9% for PV and wind prediction for the whole year respectively.