By Nidhi DhullReviewed by Susha Cheriyedath, M.Sc.Sep 26 2024A recent article published in Sustainability proposed an artificial intelligence (AI)-based framework involving Large Language Models (LLMs) of OpenAI ChatGPT-3 and Google Bard to examine 26 criteria representing smart building (SB) services across five smart cities (SCs) infrastructure domains (mobility, energy, water, waste management, and security).
Study: An AI-Based Evaluation Framework for Smart Building Integration into Smart City. Image Credit: majcot/Shutterstock.com
Background
SBs and SCs are complex, with several diverse technologies and systems operating harmoniously. Several methods are employed for SB evaluation, including the Smart Readiness Indicator and the SPIRETM SB rating program. These frameworks help stakeholders understand and improve resilience and urban sustainability using evaluation parameters like social welfare, economic efficiency, and environmental excellence.
However, existing frameworks have limitations such as difficult standardization, data access, data management, and data correctness. Moreover, only a few studies analyze the interconnection between SB and SC performance. Overcoming these challenges requires constant improvement in data infrastructure and assessment processes.
LLMs based on deep learning methods offer efficient sentiment and contextual analysis. Such sophisticated AI tools can provide better insights into SBs’ impact on smart city performance. Thus, this study developed and validated a comprehensive method to evaluate the integration of SB services into the larger SC ecosystem and infrastructure.
Methods
Three primary aspects—efficiency, resilience, and environmental sustainability—were selected to evaluate SCs' performance. Additionally, the complex interdependencies between SBs and SCs were examined across five domains: energy, mobility, water, waste management, and security. Furthermore, 26 intelligent service variables across these five SC domains were used to describe an SB's smart features.
Two well-known LLMs, Google’s Bard and OpenAI ChatGPT-3, were considered for developing the evaluation framework, which assessed the impact of selected SB services on SC performance. Accordingly, each factor in a domain was evaluated for efficiency, resilience, and environmental sustainability.
The impact of each SB factor was quantified on a scale of 0 (no impact) to 2 (significant impact). Alternatively, each SC infrastructure domain was allocated a weight from 1 (not important) to 5 (significantly important) using the 5-point Likert scale.
To mitigate biases, each AI model was repeatedly tested in five sessions of the same tasks, each with five trials. The response-generated distribution from the mean values of each session was then analyzed to determine the possible impact scores. Furthermore, the results were validated through the Delphi method with two rounds (preliminary examination and reevaluation) of a validation process.
Finally, the applicability of the developed AI framework was demonstrated through case studies on five sophisticated SB projects: The Edge in Amsterdam, One Angel Square in Manchester, National University of Singapore (NUS), Ongos Valley in Namibia, and Reliance Modern Economic Township (MET) City in India.
Results and Discussion
ChatGPT -3 consistently assigned a score of 2 to the efficiency and resilience of SC infrastructure factors. However, the score for environmental sustainability varied among these SC domains. Alternatively, Google Bard demonstrated a broader score range across all three SC performance parameters, attributed to its real-time training through Google search.
During Round 1 of result validation, a consensus of 80% was achieved for only 17 of the 26 attributes and two domains. Alternatively, an 80% consensus level was reached for the remaining nine variables and their respective importance domains after Round 2. Thus, the Delphi technique improved the AI-generated evaluation framework by utilizing iterative expert consensus.
SB services for the SC energy infrastructure domain were the most important within the framework, achieving a maximum impact of 32.67%. The second most prioritized domain was smart water management services, with a 23.96% impact on SB integration into SC, followed by the security domain, with a 19.96% impact. Alternatively, smart mobility services, crucial to mitigating SC traffic congestion, had a 17.42% impact. Finally, smart waste management services represented a minimum of 5.99% impact on the total integration into SC.
The selected case studies on the SB integration in SC demonstrated the maximum score for The Edge in Amsterdam, notably in the energy, water, and security domains. Alternatively, Ongos Valley performed exceptionally well in waste management and security domains. However, NUS, One Angel Square, and Reliance MET City exhibited some deficiencies compared to the ideal model despite notable levels of integration.
Conclusion
Overall, the researchers comprehensively evaluated the impacts of SB services on SC performance using advanced AI tools. Apart from determining the weightage of the influence of different SB services, the proposed framework was successfully applied to analyze how SB services correspond with SC aims of efficiency, resilience, and environmental sustainability through five case studies.
However, this study focused only on the technological aspects of integrating SBs into SCs. The human factors, long-term sustainability, user acceptance, and cost-effectiveness of SB solutions are important for successful SC integration. Thus, the researchers suggest expanding the framework’s application to a broader range of SB initiatives to improve decision-making on infrastructure development, resource allocation, and sustainability measures.
Journal Reference
Shahrabani, M. M. N. & Apanaviciene, R. (2024). An AI-Based Evaluation Framework for Smart Building Integration into Smart City. Sustainability, 16(18), 8032. DOI: 10.3390/su16188032, https://www.mdpi.com/2071-1050/16/18/8032
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.