This paper presents a digital model of an intelligent solar-powered street lighting system with integrated energy consumption forecasting, developed in MATLAB/Simulink. The simulation uses real hourly weather data from the Almaty region, where solar irradiance ranged from 0 to 980 W/m² and temperatures varied from –25°C to +35°C. The modeled system includes a 200 W photovoltaic panel, a 120 Ah battery, and LED luminaires up to 60 W. Predictive analysis of the battery’s state of charge (SOC) and need for external power was conducted using AI algorithms ANN, LSTM, GRU, and Random Forest trained on synthetic data from the simulation. Results show the system can operate autonomously for up to 72 hours under adverse weather. The probability of switching to backup power is 27–32% in winter and under 8% in summer. The LSTM and GRU models achieved a mean SOC prediction error of less than 5% versus actual values. The proposed architecture offers a practical and adaptable approach for designing, testing, and optimizing solar-powered lighting in both urban and rural settings across Kazakhstan and similar regions. It demonstrates the potential of combining simulation with AI to support sustainable and resilient outdoor lighting infrastructure.
Junisbekov et al. (Wed,) studied this question.