Abstract: A lightweight Extreme Learning Machine (ELM) tailored for Microcontroller Units (MCU) has been proposed, with a specific application in lithium battery State of Health (SOH) prediction. By analyzing data on lithium battery aging, this study introduces several enhancements to the conventional ELM, aimed at enhancing its computational efficiency and minimizing hardware resource consumption. Utilizing the NASA dataset, extensive experiments were conducted across different battery types, multiple programming languages, and a variety of MCUs. The results demonstrate that our proposed methodology is capable of delivering highly accurate and efficient lithium battery SOH estimations.