Photovoltaic Power Prediction Based on DBSCAN-PCA and Improved Self-attention Mechanism
The power data of photovoltaic(PV)power stations has the characteristic of randomness and volatility.Therefore,the accurate power prediction models for PV power stations are essiential for flexible power dispatching and management.This paper proposes a hybrid technology PV power prediction model based on DBSCAN cluste-ring,PCA principal component analysis and improved self-attention mechanism.Firstly,the DBSCAN clustering is used to classify the scattered and dense data to obtain the fluctuation data set and the stationary data set.Secondly,the PCA is used to extract the main component sequence of the fluctuation data to obtain the time series that is con-venient for the model to obtain key information.Finally,the key meteorological parameters are extracted to combine with the improved self attention mechanism model with perceptual context information for interactive dynamic model-ing.The comparative experiment shows that the stationary data and fluctuation data of the proposed model in each season have high prediction accuracy.
DBSCAN clusteringPCA analysisself-attention mechanismphotovoltaic power prediction
万方数据
维普
