Predicting Parameters Affecting Building Energy Consumption Using Machine Learning Models

Abstract

The rapid growth of population and the construction industry have led to an increase in energy demand and an increased pressure on the environment in urban areas, and especially in Tehran, where buildings are responsible for more than 40% of the total energy consumption of the country, due mostly to heating and cooling. Accurate estimating of buildings energy demand is dependent on the local climate because this will help tremendously in the management of energy use. This study aims to assess the predictive power of four machine learning and deep learning (DL) models: multilayer perceptron, long short-term memory (LSTM), convolutional neural network (CNN), and a hybrid CNN–LSTM model to forecast climatic parameters and subsequently use this to estimate building energy consumption. The models were trained using hourly weather data (2001–2020) and simulated energy usage data from a typical office building. Key climatic parameters influencing energy demand were identified as temperature, humidity, wind speed, and solar radiation. The study aimed to capture the complex, nonlinear, and time-dependent relationships between these variables and energy consumption. Results show that the hybrid CNN–LSTM model outperforms other models, offering superior accuracy in forecasting climatic conditions and energy demand patterns by effectively capturing both spatial and temporal dependencies. The study demonstrates the substantial advantage of DL in improving predictive accuracy and optimizing energy efficiency within building energy management systems. This research provides a robust, data-driven framework for climate-responsive energy forecasting in Tehran, offering valuable insights for policymakers and urban planners seeking to reduce environmental impacts and promote sustainable urban development.