A piezoelectric energy harvesting system with maximum power point tracking based on an artificial neural network

This paper presents an efficient piezoelectric energy harvesting system that utilises an Artificial Neural Network (ANN) for Maximum Power Point Tracking (MPPT). The system integrates a Standard Energy Harvesting (SEH) circuit with an ANN-based controller and a DC–DC buck converter to achieve dynamic impedance matching under varying vibrations. An offline-trained Multilayer Perceptron ANN predicts the optimal load-side power. A PID controller then drives the converter based on the prediction error, enabling continuous tracking without open-circuit interruption. Simulation results show a peak MPPT efficiency of 96.5% and a 57.1% power enhancement during maximum power drops. The system maintains over 92.7% efficiency for input currents of 0.2–13 mA and over 95% for loads of 100 kΩ–5 MΩ. Robustness is confirmed by an efficiency above 86.6% across an extended load range of 60 kΩ to 10 MΩ.