Enhancing Energy Efficiency in Smart Cities: Advanced ANN and Decision Tree Model for Solar Energy with IoT and Cloud Server

Abstract

In the current era of climate change challenges, employing data analysis and Machine Learning (ML) techniques has become crucial in generating precise forecasts to optimize energy consumption. Efficiency factors like solar irradiance, clear skies, clean panels, and unobstructed sun exposure play crucial roles. Using sensors for monitoring and IoT for remote accessibility, solar power systems incorporate smart technology connected to Arduinos for continuous panel parameter monitoring. This research predominantly focuses on Energy Load Forecasting in the smart city environment, developing and comparing hybrid deep learning model using historical data from a near Zero Energy Building. The dataset comprises energy load and temperature metrics, shaping various ML algorithms such as Artificial Neural Networks and Decision-trees, tailored to unique data features like the presence of photovoltaics. A groundbreaking hybrid model, amalgamating outputs from multiple ML algorithms, has been introduced, resulting in a meta-model voting regressor that standardizes new data inputs. Experimental evaluations against unseen data and alternative ensemble methods displayed promising forecasting results, exhibiting superior performance compared to base algorithms.