Power & Energy Grid Management for AI-enabled Optimal Planning
Pilot Site: AYE (Spain) & PPC (Greece)
Techonlogy Providers: AYE, PPC, SID, CERTH, TCD, UBITECH, MINDS
Innovation Providers: SID, AYE, CERTH, UBITECH, AYE, MINDS
Pilot #1 focuses on energy grid management for AI-enabled optimal planning. The pilot involves multiple actors, including AYE and PPC as pilot sites, as well as technology and innovation providers such as SID, CERTH, TCD, UBITECH, and MINDS. The rationale behind the pilot is the increasing integration of renewable energy sources into energy grids, which presents challenges in terms of planning due to the different characteristics of these sources.
The pilot addresses the energy management problem from both the supplier and consumer perspectives, utilizing AI models for energy generation and demand/supply forecasting. The pilot aims to train predictive models using big data generated from sensors, devices, and appliances, and to optimize these models using the RAIDO platform. The impact of the pilot is to increase the efficiency and resilience of the energy system, reduce CO2 emissions, and demonstrate the integration of XAI and game theoretic techniques in AI training and deployment. The setup involves interconnected power plants, power grids, and smart homes, with edge computing units hosting the EMS and AI models.
The pilot will take place in AYESA, PPC, and CERTH premises, and will involve steps such as collecting historic data, retraining and optimizing AI models, designing lifelong models for adaptation, and evaluating the optimized models. The pilot will utilize Digital Twins and simulated data for training and optimization, and will employ blockchain for continuous feedback and monitoring.
Autonomous & AI Powered Monitoring & Operations
Pilot Site: HELD/KRE (UK/Greece, Belgium)
Techonlogy Providers: HELD, KRE, 8BELLS, TCD, UBITECH, QMUL
Innovation Providers: KU, SID, QMUL
Pilot #2 focuses on precision agriculture in the domains of autonomous and AI-powered monitoring and operations. The pilot site is located in the UK, Greece, and Belgium, with key technology providers including HELD, KRE, 8BELLS, TCD, UBITECH, and QMUL.
The pilot aims to address two scenarios: pharmaceutical cannabis cultivation and fungal processes for meat replacement. For pharmaceutical cannabis cultivation, the pilot aims to use multispectral photography and AI training models to detect diseases early and accurately identify the optimal harvesting time. In the case of fungal processes, the pilot aims to use timeseries data and images to predict nutrition, viscosity, and color parameters and monitor the age of the fungi to ensure high-quality end products. The pilot requires big volumes of data from sensors and cameras, and the use of simulated data to reduce the number of sensors needed. The impact of the pilot includes increasing indoor farming sustainability, reducing energy consumption, and optimizing yield and quality.
The setup involves IoT devices with RGB-D cameras and multispectral sensors, and the deployment of optimized AI models. The demonstration will take place in Greece and Belgium, utilizing historical data to build digital twins of smart farms and training AI models for disease detection and harvesting time prediction. The models will be optimized and evaluated using the RAIDO platform.
Digital Health Solutions for Personalised Preventive Pharmacogenetics
Pilot Site: VITO, JESSA (Belgium)
Techonlogy Providers: VITO, TCD, SID, 8BELLS
Innovation Providers: VITO, KU, QMUL, SID
Pilot #3 focuses on implementing an AI system with a human-in-the-loop strategy to design personalized preventive pharmacogenetics (PGx) application archetypes for the healthcare domain. The goal is to address challenges such as diversity in actionable PGx information, differences in socio-economic status and education level, varying understanding of PGx information by healthcare professionals, and the need for precise and dynamic PGx information. The pilot aims to improve communication, understanding, and implementation of PGx information through AI models, bias detection, feedback, and monitoring processes enabled by blockchain technology. The impact of this pilot includes improved communication, reduced unfairness, increased clarity of information, and enhanced safety and efficacy of drugs prescribed in primary healthcare. The setup involves generating and augmenting data from various sources, optimizing AI models for performance and interpretability, and involving patients and healthcare professionals in the validation process. The demonstration will evaluate the PGx application with medical professionals in real-life settings, focusing on the analysis of potential patients, communication of results and reports, and decision-making by the doctors. The AI models will be trained and fine-tuned using real and synthetic data, with a focus on summarization, ethics, bias detection, and explainability. The pilot will also collect feedback through questionnaires and interviews to improve the application and its integration into the workflow of medical professionals
Industry 5.0 & Bio-based Composites, AI Models for Plant Fibre
Pilot Site: UBFC (France);
Techonlogy Providers: UBITECH, TCD, UBFC, SID, 8BELLS
Innovation Providers: UBFC, MINDS, 8BELLS, SID
Pilot #4 focuses on the characterization of plant fibers for use in bio-based composites and robotics. The pilot aims to address the challenges of high cost and limited properties of fully biodegradable materials by developing Plant Fiber Reinforced Composites (PFRCs) as an alternative to conventional composites. However, the mechanical characterization of plant fibers is difficult and time-consuming. The pilot proposes using advanced deep reinforcement learning (RL) architectures with human-in-the-loop (HITL) to solve this problem. The pilot will implement ML/DL pipelines to accelerate simulation-based mechanical property estimation and predict the mechanical properties of fibers directly from raw data. The goal is to enable fast and reliable characterization of plant fibers, which will facilitate the design and production of PFRCs and contribute to the scaling-up of bio-based composites manufacturing. The pilot will be demonstrated in UBFC (France) and will involve three phases: data generation/collection, AI V-L models design and training, and optimization and explainable AI (XAI). The pilot aims to achieve a 15x faster mechanical property estimation with over 95% accuracy compared to current methods
