Novel load-participated all-time multivariate load prediction method

The problems that a traditional planning load prediction model is simple and single in element, and source load storage resources are fully considered are solved, a novel load-participated full-time-period multi-element load prediction method is provided, and through models and algorithms based on a GRU-GAN model, a 2DGAR-BiLSTM model, a DBSCAN algorithm, a multi-element load characteristic load flow calculation optimization algorithm-K-means clustering algorithm and the like, the load prediction efficiency is improved, and the load prediction efficiency is improved. Load characteristics such as randomness, time-varying characteristics and volatility of multi-element loads are fully considered, theoretical data and load curves are combined, and targeted and refined load prediction functions such as distributed photovoltaic output prediction, charging pile load prediction, energy storage demand analysis and conventional load prediction optimization are realized through data analysis and environment variable an