Using deep learning to design an intelligent controller for street lighting and power consumption
Bilal Ibrahim Bakri,Yaser M. Abid,Ghaidaa Ahmed Ali,Mohammed Salih Mahdi,Alaa Hamza Omran,Mustafa Musa Jaber,Mustafa A. Jalil,Roula AJ. Kadhim
DOI: https://doi.org/10.15587/1729-4061.2022.260077
2022-06-30
Eastern-European Journal of Enterprise Technologies
Abstract:Street lighting is very important now-days especially at dangerous areas and highways but it consume a lot of power and it became challenging for many researchers in the past few years. Enormous efforts have been placed on the issue of reducing power consumption in illuminating cities and streets, researchers had various approaches and methods in tackling this challenging matter, till now there is no ideal system that has been developed to reduce the electricity usage. In this paper intelligent controller based on deep learning proposed to control the light at the street from sunset to sunrise, the system will decrease the light used to illuminate the streets in the absence of movements, the network trained based on deep learning with several image of different objects to help the system detecting any moving objects in the street to provide the street with the exact amount of light needed in order to reduce the waste of electrical energy resulting from street lighting and to help reduce accidents hence high percentage of criminal activity and life threatening conditions occur in the absence of light. The system was trained with a vast and diverse dataset to assure the accuracy and efficiency of the proposed system, the trained system showed a result of 90 precision of detecting moving objects, the proposed system was tested with a new dataset to assure the reliability and dependency of the system and reducing the errors to the minimum, the system shows promising results in detecting movements and objects, after the detection being complete, the system will send a pulse width modulation causing a 20 % light dimming, leading to enormous reduction in the power consumption, adding to that the proposed system is easy to use