<![CDATA[The hybrid model for very short-term photovoltaic (PV) power forecasting, covering one hour ahead with 20-minute intervals, combines the k-nearest neighbour (k-NN) and multilayer backpropagation neural network (BP-NN) methods. The uniqueness of this model lies in integrating meteorological and the clarity index. the data preprocessing stage, the k-NN method is applied, while the multilayer BP-NN is used for forecasting. The k-NN Multilayer BP-NN algorithm calculates the nearest data points using Euclidean distance, and then processes the training and testing data through the multilayer BP-NN to generate PV power predictions. The simulation dataset was divided into 70% training data and 30% testing data, with a maximum PV power output of 611 W. The error statistical indicators of machine learning using k-NN-BP-NN model RMSE 27.44 W and MSE 1.5 W. These superior results are attributed to more stable weather patterns and consistent solar radiation. The simulation validity test demonstrated that the k-NN Multilayer BP-NN algorithm achieved better accuracy compared to the k-NN decomposition method. In addition, the model offers high computational efficiency and short inference time, making it highly suitable for real-time PV power forecasting systems.]]>
Copyrights © 2026
