Design and Implementation of Deep Learning-Driven Sensorless Controller for DC LED Street Lights

Abstract The adoption of energy-efficient and intelligent control systems is critical for modern LED-based street lighting solutions. Traditional control methods, while effective, often struggle to adapt to non-linearities and dynamic changes in system parameters. This paper presents the design and implementation of a novel sensorless controller for boost converter driven DC LED Street lights, leveraging deep learning to achieve accurate and robust voltage regulation without a physical voltage sensor. A Deep Neural Network (DNN) is trained to estimate the output voltage of the system using measurable parameters such as input voltage, load current, and duty cycle. This estimated voltage is then utilized in a control loop to maintain stable operation under varying load and environmental conditions. The DNN model is trained on simulated and experimental data representing a wide range of operating conditions, including load variations. The controller is implemented on DSP controller platform and validated on a laboratory prototype of 500W DC boost converter-based LED driver. Performance evaluation is demonstrated using simulation and experimental results. This work provides a scalable and cost-effective solution for modern street lighting systems, paving the way for smarter and more sustainable urban lighting infrastructure.