Tallinn University of Technology

Abiodun Emmanuel Onile, the PhD student of the Department of Software Science, will defend his PhD thesis ”Innovative Energy Services based on Behavioural–Reflective–Attributes and Intelligent Recommendation Systems“ (”Arukatel soovitussüsteemidel põhinevad uuenduslikud energiateenused“) on November 12, 2024, starting at 13:00 . The defense will take place in room ICT-638 (Akadeemia tee 15a, ICT building of TalTech) and can be also followed via Zoom.

In recent times, the escalating impact of urban climate change and the increasing demand for electrical energy have become pressing concerns. Broadly, electricity conservation strategies rely on efficient utilization. Yet, challenges arise due to consumers facing difficulties in executing these measures. In order to achieve the desired efficiency in energy usage, innovative solutions aimed at influencing the behavior of energy consumers are necessary. Personalized recommendations play a crucial role in stimulating behavioral changes among consumers, thereby contributing to sustainable advancements in energy efficiency. Addressing this challenge, this thesis centers on innovative energy services reliant on intelligent recommendation systems and digital twins. The thesis ”Innovative Energy Services based on Behavioural–Reflective–Attributes and Intelligent Recommendation Systems“ examines various trends related to modeling and the adoption of energy services, considering the positive connections between recommendation systems and the energy behaviour of consumers on the demand side. Through a comprehensive content analysis of leading research works in the IEEE Xplore and Scopus databases, our study validates the innovative use of data-driven twin technologies for demand-side recommendation services.

Industry 4.0 plays a role in end-consumers progression towards energy efficiency by enabling more sophisticated analytics and creating avenues for end-consumers and decentralized grid assets to be modeled similarly to their Digital Twin (DT) counterparts. This development opens pathways for asset-level analytics. This study introduces an innovative approach employing an ensemble of hybrid DT asset models, combining ordinary differential equation (ODE) physics engines and data-driven recurrent neural network (RNN) prediction methods. The emerging real-time information technology (IT) applications and innovative modeling of DT for individual electricity assets are disrupting this landscape. Particularly, in relation to integration of intermittent renewable energy sources which offers promising demand-response (DR) solutions, but also brings about emerging stability issues for the electricity grid. Similarly, aggregators are increasingly pivotal in the DR electricity market, yet they frequently grapple with substantial market monopolization and a lack of transparency by profiteering or taking advantage of transitive RES. Additionally, following modelling of individual asset based predictive DT, this study introduces an novel graph-based ranking approach using PageRank model for demand side recommendation service provision.

There are specific scenarios where end-users might disregard recommended advice, contributing to a widening ’knowledge-action gap’. A demand-side recommender system, complemented by Generative Pre-trained Transformers (GPT) for a conversational chatbot interface technology and Extended Reality (XR) proves effective in engaging and extending end-consumer interest in recommended advice. The novelty of this research lies in developing a new approach that enables individual end-user assets to contribute to demand-response initiatives.

Results from the study demonstrate that employing BESS through the multi-agent reinforcement learning control strategy yielded a maximum peak load reduction of approximately 24.5%, alongside a 94% and 69% improvement in comfort for specific loads and users’ engagement respectively. Furthermore, efficiency-related recommendations for BESS contributed to a linear reduction in peak load, surpassing the baseline scenario. The study’s outcomes also show that the hybrid Digital Twin approach accomplished an impressive 84.132% reduction in prediction error, as measured by MSE metrics.

The thesis ”Innovative Energy Services based on Behavioural–Reflective–Attributes and Intelligent Recommendation Systems“ is published in the Digital Collection of TalTech Library.

Supervisor Prof. Juri Belikov and co-supervisor Prof. Eduard Petlenkov.

Oponents:

  1. Prof. Bo Nørregaard Jørgensen, SDU Center for Energy Informatics, University of Southern Denmark, Denmark;
  2. Prof. Lucio Tommaso de Paolis, Department of Engineering for Innovation, University of Salento, Italy.

Follow public defence in Zoom

Meeting ID: 961 7961 9964
Passcode: 914910