Presentation Information
[A-8-12]Real-time Learning-based Power Management for Energy Harvesting Systems Using A Fixed-point Extreme Learning Machine
◎Enming Zhang1, Takeaki YAJIMA1 (1. Kyushu University)
Keywords:
MPPT,ELM-RLS,Energy Harvesting,16-bit Fixed-Point,IoT
The rapid growth of the Internet of Things (IoT) has increased the demand for efficient Maximum Power Point Tracking (MPPT) in energy harvesting systems. This paper proposes a 16-bit fixed-point MPPT method based on an Extreme Learning Machine with Recursive Least Squares (ELM-RLS). The proposed method predicts the output power under different operating conditions and selects the optimal switching frequency for real-time MPPT. Experimental data collected from a vibration energy harvesting system are used for training and evaluation. The results demonstrate that the proposed 16-bit implementation achieves prediction accuracy comparable to that of the double-precision model while significantly reducing computational complexity and estimated energy consumption. These characteristics make the proposed method suitable for practical low-power embedded energy harvesting systems.