Research on Improving Accuracy of Machine Learning Power Prediction System Based on Semi-Supervised Strategy

The accuracy of wind and solar renewable energy power forecasting not only impacts the assessment of “dual guidelines” for each wind and solar energy site, but also affects the electricity market-based transactions and power generation revenue of these facilities. Employing deep learning models, we amalgamate meteorological grid characteristics, temporal features, weather pattern scenarios, and refined wind speed and power observation data. Utilizing machine learning techniques, we integrate diverse numerical forecasting outcomes and refine wind speed predictions. Furthermore, transitioning from wind to electricity models, we directly forecast wind power outcomes, either through wind speed adjustments or directly generating wind power predictions. In summation, based on the geographical location, meteorological attributes, node consumption conditions, frequency, and types of extreme weather events in the current subsidiary's renewable energy site distribution area, we undertake the research and development of a high-precision power prediction and electricity generation estimation system rooted in a semi-supervised strategic machine learning framework.