AI and Big Data Revolutionizing Low-Carbon Buildings: Challenges and Promises

By Dr Silpaja Chandrasekar, PhDReviewed by Susha Cheriyedath, M.Sc.Aug 15 2023

In a paper published in the journal Sustainability, researchers highlighted the significance of mitigating carbon emissions from buildings in attaining global carbon neutrality targets. It provides an overview of the current state-of-the-art methods involving AI and big data for building energy conservation and low-carbon practices.

Background

The growing concern over global warming due to carbon emissions has led to urgent calls for action. The "Paris Agreement" strives to limit global temperature increases, with various countries aiming for carbon neutrality by 2050. Amid rapid economic growth and urbanization, carbon emissions have surged, necessitating immediate reductions. The construction sector, responsible for a substantial portion of global emissions, is a key focus.

The building sector faces various challenges, including inaccurate forecasting, inadequate measurement methods, and inefficient accounting for energy usage and emission reduction. Recent studies have indicated that leveraging artificial intelligence (AI) and big data technologies could substantially enhance the precision for predicting building energy usage. These results could subsequently be utilized for effective building operation management, thereby contributing to emission reduction objectives.

Buildings account for around 39% of global CO₂ emissions, prompting efforts to implement sustainable practices. In this context, the present study examines the role of AI in revolutionizing construction for emission reduction and low-carbon operations. It also explores carbon accounting approaches and proposes ML solutions to drive environmentally-conscious practices.

AI and Big Data for Building Energy Efficiency

Exploring key building energy prediction aspects, the study begins with data preprocessing and advances to utilizing neural networks and deep learning for low-carbon buildings. AI and big data drive data analysis, assisting decision-making to reduce carbon emissions for administrators and energy managers. Amid intricate building energy consumption data, adept preprocessing tackles challenges. However, the lack of interdisciplinary research raises emission concerns, and data selection and preparation remain vital challenges as neural networks require reliable data for dependable decisions.

The study categorizes preprocessing into integration, cleaning, reduction, and transformation, using techniques like K-means clustering and ML for data filling. The emphasis on robust modeling data highlights the crucial role of data quality in achieving accurate predictions and informed decision-making.

Recent research extensively explores modeling for precise building energy prediction. From autoregressive models to attention mechanisms, researchers enhance accuracy. TimeGrad presents a probabilistic time series approach outperforming existing models. Attention mechanisms, inspired by Transformers, compare features. Hybrid models, combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory networks (LSTMs), enhance spatial-temporal dependencies. Incorporating COVID-19 data improves accuracy, signaling a shift toward precise energy management in buildings.

Machine Learning for Low Carbon Buildings

ML is crucial in reducing building energy consumption, encompassing power grid control, reinforcement learning for optimization, power market dynamics, and electric vehicle integration. These applications optimize energy usage, curb carbon emissions, and target building carbon reduction measures. Moreover, smart grids and integrated energy systems are harnessed for sustainability, aided by reinforcement learning. ML enhances electricity market strategies and electric vehicle charging efficiency, though intermittent control strategies pose challenges for certain applications.

Carbon Emission Accounting Empowered by Big Data

Carbon emissions are tracked through three main methods: emission factor, mass balance, and actual measurement. Building carbon accounting uses top-down (macro level) and bottom-up (micro level) methods. Bottom-up involves detailed consumption calculations, while top-down provides overall estimates. Both have advantages and drawbacks. Regional building accounting typically employs top-down, focusing on macroeconomics. Conversely, bottom-up considers the life cycle stages of the building, including materials and operation. Single building accounting follows lifecycle phases from design to demolition. While the challenges include emission boundary definition and lack of standardization, effective carbon reduction requires addressing upcoming emission prediction and component-level analysis.

Obstacles and Outlook

The major challenges in building carbon emissions involve improving ML prediction accuracy under real-world scenarios while safeguarding privacy and addressing inconsistent standards. Solutions include refining building features, enhancing data quality, and utilizing privacy-preserving technologies like federated learning and differential privacy. Integrating Internet of Things (IoT) and big data can establish a monitoring platform for real-time carbon evaluation, aiding emission reduction strategy formulation. Visualization tools like histograms and interactive maps help assess the effectiveness of reduction measures and regional impact. Creating a big data platform for building operation data enhances comprehensive data collection, automates calculations, and supports sustainable development by identifying energy waste and emission reduction opportunities.

Contributions of this Paper

The key contributions of this study are the following:

• Various AI techniques like LSTM, Generalized Regression Neural Network (GRNN), and Artificial Neural Network (ANN) for emission prediction are compared in this study, along with potential uses for emission reduction.
• Preprocessing methods are discussed for carbon data, including K-nearest neighbors (KNN) for filling gaps, anomaly detection, and dimension reduction.
• Carbon accounting models for buildings at regional and individual levels are covered, including China Building Carbon Emission Model (CBCEM).
• A big data platform for monitoring building operations is recommended.
• Privacy protection and positioning are suggested for future research.

Conclusion

To summarize, this article explores construction data preprocessing techniques like decomposition, clarification, and augmentation. It outlines ML methods for energy prediction and examines the potential of AI in carbon emissions reduction. Additionally, it explains building life cycle theory and carbon accounting for individual and regional levels.

 The authors highlighted the urgency of carbon emission reduction in the construction industry, which plays a substantial role in global greenhouse gas emissions. With a focus on AI-driven methods for emission reduction and low-carbon operations, the research proposes ML solutions to support environmentally friendly practices in the construction sector. As AI and big data continue to shape the construction landscape, this paper aims to contribute to the ongoing efforts in improving building energy conservation and advancing sustainable practices, underscoring the potential for positive impact in this critical area.