Decentralized control system for unlimited street lighting poles with an intelligent, energy-saving off-grid maximum power point tracking battery charger

• Design and Implementation: Introducing a novel scheme for a decentralized street lighting system capable of managing an unlimited number of light poles. • Intelligent MPP Tracking: Proposing an intelligent MPP tracker based on deep ANN algorithm to ensure an efficient off-grid PV system, maximizing power harvested from incident sunlight. • Battery Charging Process: Proposing a battery charging strategy that monitors the state of charge (SOC) and terminal voltage, ensuring charging remains within the recommended minimum and maximum levels. In an attempt to tackle the critical issue of CO 2 emissions and embrace sustainability, we propose an energy-efficient street lighting system. The system offers a novel, adaptive, and decentralized dimming solution for an unlimited number of Light-Emitting Diode (LED) streetlight poles, which dynamically adjusts the brightness in response to traffic and pedestrian presence. Each light pole employs motion detectors to detect street occupancy, allowing responsive illumination that improves energy efficiency without compromising safety. To demonstrate this idea, a prototype consisting of four LED light poles was developed. Additionally, we investigate how solar energy as a clean renewable source might be included in the system, offering an off-grid street lighting dimming solution. A deep artificial neural network (ANN) algorithm is designed to have an effective response of maximum power point tracking (MPPT) in terms of accuracy and speed to obtain maximum electrical power from the incident light on a pair of photovoltaic panels fixed above an off-grid street light pole. We gathered and examined data on the poles' electrical use, demonstrating the system's efficiency in cutting down on energy use, and it is concluded that the percentage of energy saving by the proposed system is 49.22% compared to the full lighting system. Simulation results in MATLAB/Simulink show an efficient deep ANN algorithm with a mean squared error (MSE) of 3.157 × 10 -5 at epoch 116 and a flexible setting to State of Charge (SOC%) based on battery specifications limits, demonstrating a steady battery charging voltage.