Short-term PV Power Prediction based on VMD-SE-TCAN Model

Short-term PV Power Prediction based on VMD-SE-TCAN Model

Focusing on the accuracy of short-term photovoltaic power prediction, this paper proposes a VMD-SE-TCAN short-term photovoltaic prediction model. First, the Sample Entropy (SE) is used to fuse the different frequency components generated by the Variational Modal Decomposition (VMD) to reduce  the complexity of prediction model and effectively alleviate the under-decomposition or over-decomposition of traditional decomposition algorithm; then the TCAN model based on Temporal Convolution Neural network (TCN) and Attention Mechanism is used to predict each component, the TCN is used to capture temporal dependencies in prediction, and the Attention Mechanism improves the impact of key weather features on different components; finally, the prediction results of each component are superimposed to output photovoltaic prediction power. Through experiments and comparison verification, the combined prediction method effectively improves the accuracy of photovoltaic power prediction, and can better guarantee the reliable operation of the power system.