Two-Stage Short-term Power Prediction of Small Hydropower Station Based on Dimension-Reduction Clustering and Hybrid Neural Network

As a clean, renewable and adjustable energy source, small hydropower plays an important role in solving the problem of power supply in remote areas, facilitating the low-carbon energy structure, and promoting the consumption of wind and photovoltaic power. Accurate short-term power forecasting for small hydropower helps to balance power generation and load of the system, and ensure the safe and stable operation of the power grid. In this paper, a two-stage short-term power prediction method for small hydropower stations based on dimension-reduction clustering and hybrid neural network is proposed. Firstly, the correlation between small hydropower generation and various external factors is analyzed by Pearson correlation coefficient (PCC), and the dimension-reduction of power data is done by using sparse principal component analysis (SPCA); then the fuzzy C-means (FCM) algorithm is used to classify the data; finally, the hybrid neural network model based on CNN-BiLSTM is used to make a two-stage prediction to improve the prediction accuracy, where the first stage selects a similar classified dataset based on the original data, and the second stage gets a better prediction based on the classified data. Case studies demonstrate that the proposed model has higher prediction accuracy and better adaptability than traditional neural networks.