A Short Term Wind-Photovoltaic-Load Forecasting Model Based on Coupling Relationship and Multi-Layer Extraction Multi-Task Learning

The integration of renewable energy into the grid has deepened the uncertainty of the power system and brought difficulties to the economic dispatch of the power system. A better response is to accurately predict the Wind power, Photovoltaic power and Load power, which can provide effective data support for the economic dispatch of the power system. To this end, we propose a short-term forecast method for Wind, Photovoltaic and Load based on coupling feature mining and Multi Task Learning (MTL). Firstly, analyze and extract the correlation between the three predicted objects, and then use a Multi Layers Extraction MTL neural network as the forecast algorithm to further represent the coupling relationship between the predicted objects; Finally, a set of Wind, Photovoltaic and Load data was used as experimental data to compare and analyze the predicted results.