Aug 2026· 2026 6th International Conference on Emerging Smart Technologies and Applications (eSmarTA)· pp. 1-6· 0 citations· 11 references
Abstract
HVAC systems use up about half of the total energy in smart buildings and are a key focus of optimization. The demand of HVAC energy is very difficult to forecast with high accuracy due to the nonlinear nature of HVAC operations, high temporal variability, and interdependencies among environmental and operational variables. Traditional forecasting methods like regression based models and ARIMA often do not reflect such multivariate dependencies resulting in incompetent energy management. This paper presents a multivariate Long Short-Term Memory (LSTM) model that will be developed to learn the long-term temporal dynamics of various variables related to HVAC. The model is trained and tested on a real-world benchmark dataset, which includes 11 sensor-derived features, and uses one fully connected LSTM layer with 50 hidden units trained using the Adam algorithm. Root Mean Square Error (RMSE) and the coefficient of determination (R2) are reported per variable as measures of forecast performance. The experimental findings indicate that the model is accurate, over 90% on most variables, a fact that justifies the fact that the model is effective in overcoming the weaknesses of the traditional methods and giving accurate predictions that can be incorporated into smart building energy management systems. Further research will focus on hybrid deep learning networks and TinyML networks to run on edge devices that are IoT-enabled.
Global energy consumption has increased significantly in recent decades-rapid urbanization and technological advancement. That means sustainability has become a critical global concern; it's something we need to figure out, fast. Buildings account for significant proportion of energy consumption, especially for heating, ventilation, air conditioning and appliances. Smart green Buildings have emerged as a promising solution, but honestly, most management systems running them still use conventional approaches that lack adaptability when the weather shifts or the number of people inside changes. This study addresses this limitations. We built an AI-powered system to predict and model energy use in these smart green buildings. Several machine learning models were implemented-Linear Regression, Random Forest Regression, and LSTM networks-analyzing historical energy consumption data along with environmental info like temperature, humidity, and time of day. To see which model actually works, we evaluated model performance with Mean Absolute Error, Root Mean Square Error, and ${R}^{{2}}$ scores. Turns out, deep learning models and ensemble approaches like LSTM and Random Forest significantly outperform traditional linear regression models, especially when the energy usage gets weird, nonlinear, or changes over time. At the end of the day, we're aiming to give assist facility managers in decision making: predicting how much energy they'll use, reducing energy waste, and making the whole setup more sustainable. This contributes to development of smart infrastructure, and honestly, that's exactly what we need right now.
Naman Verma, Vaishali Dixit· 2026 11th International Conf...· 0 citations
The increasing global energy needs especially in the building sector, which contributes major percentage of the total global energy use, requires smart and data-driven solutions to effective energy management. Heating, Ventilation, and Air-Conditioning (HVAC) systems represent major energy consumption of all building, and thus the precise energy consumption forecast is an essential requirement of the sustainable use of smart buildings. In the current study, a detailed machine learning (ML) model to forecast energy usage in an IoT-monitored educational facility in Jaipur, India, during one operation regimes, namely mechanical cooling (AC-ON) is provided. Distributed IoT sensors were systematically used to gather a real-world dataset of 365 daily observations to measure multivariate thermal, solar and envelope heat transfer, ventilation and energy parameters. Mechanical ventilation coupled with infiltration and dry-bulb temperature outdoors proved to be the main source of cooling energy changeability. Six regression algorithms were comparatively analyzed: Multiple Linear Regression, K-Nearest Neighbors (KNN), Support Vector Regression (SVR), Classification and Regression Tree (CART), Gradient Boosting Machines (GBM), and Extreme Gradient Boosting (XGBoost). Mean Absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R 2) were used to evaluate the model performance. Ensemble models were significantly improved predictors; Gradient Boosting had the minimum prediction error in cooling energy (MAE: 6.420, R 2: 0.990) and equipment energy consumption (MAE: 0.2716, R 2: 0.9997). The analysis of the scatter plot revealed prediction diagnosis which indicated close clustering of the residues around the reference line which was ideal, which proved the model dependability at the extremes of the seasons. The results serve to legitimize ensemble machine learning as a sound premise toward the smart building environment in terms of defining intelligent energy optimization and demand-responsive control strategies.
Siddharth Gupta, Renu Bagoria· Journal of Dynamics and Cont...· 0 citations
Forecasting energy demand is critical to resource optimization, grid operation, and sustainability for smart buildings and urban energy systems. This study presents a probabilistic forecasting framework designed to jointly predict hourly electricity and heat demand for a residential building using deep learning. The model ingests hourly multivariate data spanning multiple years, comprising two energy targets — electricity and heat demand — and five weather covariates — air temperature, relative humidity, wind speed, solar irradiation, and air pressure — all of which are known to significantly influence building energy consumption. The proposed framework is based on a Long Short-Term Memory (LSTM) neural network trained using look-back windows of 48 h. Performance is evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). The proposed LSTM-based framework achieves strong forecasting accuracy, with a 1-hour horizon yielding MAE = 157.7 kW and RMSE = 218.4 kW for electricity, and MAE = 128.0 kW and RMSE = 184.8 kW for heat. Even for 720-hour forecasts, errors remain low, and the uncertainty intervals are well-calibrated (Mean Prediction Interval Width (MPIW) ≈ 308–377 kW, Prediction Interval Normalized Average Width (PINAW) ≈ 0.20–0.25). Ablation experiments highlight the significance of interpolation, normalization, cyclic encoding, and weather features, while SHapley Additive exPlanations (SHAP)-based explainability provides interpretable insights into energy demand behaviour. Comprehensive benchmark evaluation against Persistence, Linear Regression (LR), Vector Autoregression (VAR), and Quantile Regression (QR) baselines confirms superior probabilistic performance (Continuous Ranked Probability Score (CRPS) improvement of 21.6% for electricity, 22.9% for heat; Energy Score (ES) improvement of 16.1%; Variogram Score (VS) improvement of 16.8%). An empirical comparison between the proposed joint model and two independent LSTMs validates the joint architecture through a 21.4% Variogram Score improvement and 12.8% Energy Score improvement.
H. S, S. Radhakrishnan· Discover Sustainability· 0 citations
The transition toward smart manufacturing requires advanced energy management strategies that leverage artificial intelligence to improve operational efficiency and sustainability. This study proposes a novel deep learning framework based on a Long Short-Term Memory (LSTM) network for analyzing and predicting energy consumption in smart manufacturing environments using real-time data acquired from Internet of Things (IoT)-enabled industrial sensors. Unlike previous studies that primarily focus on offline energy forecasting or static datasets, the proposed approach integrates temporal energy consumption patterns from heterogeneous sensor streams to support predictive energy management and dynamic load optimization. The collected data were preprocessed through normalization and feature engineering before being trained and evaluated using the LSTM model. Experimental results demonstrate that the proposed model achieves a Mean Absolute Error (MAE) of 0.84 kWh, a Root Mean Square Error (RMSE) of 2.13 kWh, and a coefficient of determination (R²) of 0.987, indicating high prediction accuracy. Furthermore, the predictive framework enables an estimated energy consumption reduction of 14.8% through proactive load scheduling. These findings demonstrate that integrating LSTM-based deep learning with IoT sensor networks provides an effective solution for intelligent energy forecasting, improves manufacturing efficiency, and contributes to sustainable industrial development.
To improve the energy-saving effect of buildings, we need to know how much electricity will be used in advance; that is, we should be able to accurately predict the electricity demand of different branches. However, due to the irregular and complex changes in the time of electricity demand, it is very difficult to accurately forecast the cooling load. Therefore, a combination of forecasting methods was used in this study to solve this problem, and the results are as follows: First, a detailed 3D model of the building is constructed to collect all kinds of data on cooling load, and then a new method integrating Grey Wolf Optimizer (GWO), Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM) is trained and tested; this method has been named the GWO-CNN-LSTM model, and it can solve problems such as irregular timing and prediction difficulty. After training, a simple optimization algorithm is employed to automatically tune the hyper-parameters of the combined network and connects the feature extraction part with the memory function of the sequence processor; through this combination, the model can better seize not only the long-term trend but also a short-term fluctuations of load data, and it has performed reasonably well compared to previous reference models and generated slightly more accurate calculations; thus, it can be concluded that this GWO-CNN-LSTM forecasting approach is relatively reliable and can provide data support for optimising AC system operation and promoting building energy conservation.
Kunyu Liu, Chao Zhang, Xuelong Zhang et al.· Applied and Computational En...· 0 citations
Accurate forecasting of power consumption is critical for efficient energy management, grid stability, and cost reduction. This study explores the application of advanced machine learning models to predict short-term and long-term power usage patterns. By leveraging historical consumption data alongside relevant external factors such as weather conditions, time of day, and economic indicators, the proposed approach employs algorithms including Random Forest, Support Vector Machines, and Deep Learning networks. The models are trained and validated on real-world datasets to evaluate their predictive accuracy and robustness. Results demonstrate that machine learning techniques significantly improve forecasting precision compared to traditional statistical methods. This enables smarter energy distribution, better demand response strategies, and supports the integration of renewable energy sources, thereby contributing to sustainable power system operation. Traditional statistical methods often fall short in capturing the complex, non-linear patterns of modern electricity usage. This project explores advanced machine learning techniques such as Random Forest, Support Vector Machines, and LSTM networks to predict both short-term and long- term electricity demand. By leveraging historical data, weather conditions, time-based factors, and user behavior, these models demonstrate superior forecasting performance compared to conventional methods. The study shows that machine learning enables smarter load balancing, better integration of renewable energy, and improved decision-making in power distribution systems. The proposed approach supports the development of intelligent, sustainable, and data-driven energy.
Keywords: Power Consumption Forecasting, Machine Learning, Random Forest, LSTM, Smart Grid, Energy Management, Electricity Demand Prediction, Time Series Forecasting.
M. Tarani, Tothadi Sowjanya· International Scientific Jou...· 0 citations
