Comparative Study on Machine Learning Based Decision-Making for Microgrid Component Operation
摘要
Proposed energy management strategy for micro-grid (MG) with renewable energy storage (hydrogen tank) reduces emissions. Comparing k-Nearest Neighbors (k-NN) and Random Forest machine learning methods, RF is more suitable for decision making on storage system components (battery, supercapacitor, fuel cell). Ensures continuous load supply, reduces non-renewable energy consumption, and maintains storage component health. The finding results indicate that the RF method outperforms k-NN in terms of accuracy with a high accuracy of around 90%, while the k-NN method had a maximum accuracy of 61% across the four control cases.