Short-term photovoltaic power prediction study based on similar day clustering and machine learning

Aiming at the challenge of short-term prediction caused by the stochasticity and uncertainty of power fluctuations in PV power systems, this paper proposes a new method for short-term prediction of PV power based on k-means++ clustering, empirical modal decomposition of fully adaptive noise ensemble (ICEEMDAN), Improved Red-tailed Hawk Optimization Algorithm (IRTH), and Least Squares Support Vector Machine (LSSVM). First, the k-means++ algorithm is used to classify the historical data into weather types such as sunny, cloudy and rainy days; second, ICEEMDAN decomposes the raw PV power data into a number of intrinsic modal functions IMFs; IRTH optimizes the kernel and penalty parameters of the LSSVM model with the aim of solving the sensitivity problem of the traditional LSSVM in the parameter selection. The experimental results show that the prediction model proposed in this paper exhibits better prediction accuracy in the short-term prediction of PV power generation.