The User Experience as a Demand Response Action in Large All Electric Ship

This thesis investigates how user experience (UX) design and artificial intelligence can be used as explicit engineering variables to improve energy efficiency in all‑electric passenger vessels. It integrates wearable conversational AI for passengers with redesigned power‑management dashboards for the technical crew, targeting both sides of the ship’s socio‑technical system. The proposed architecture links an on‑device, edge‑optimized large language model (Gemma‑2B) with a retrieval‑augmented generation pipeline and an LLM‑based middleware agent that interfaces with the Power Management System (PMS), repeatedly monitoring current, frequency, and load balancing and exposing human‑readable states for decision support. In parallel, the PMS human–machine interface is redesigned using UX methods and comparative usability testing, reducing task execution times, interaction errors, and ambiguity in key parameters, and improving situational awareness for generator dispatch and hotel‑load supervision. Methodologically, the work combines iterative prototyping, low‑fidelity Wizard‑of‑Oz evaluation of the wearable assistant, structured usability testing on legacy and redesigned PMS dashboards, and engineering‑level scenario simulations. In a controlled trial with 20 cruise passengers and five scripted scenarios covering cabin lighting, HVAC (heating, ventilation, and air conditioning) set‑points, shower duration, standby reductions, and a gamified eco‑challenge, the smartwatch‑based assistant achieved a median System Usability Scale (SUS) score of 84/100 and a 76% acceptance rate for energy‑saving suggestions. Dashboard tests with technical operators showed consistent improvements in completion intervals and success rates for generator and load‑management tasks, indicating clearer feedback and reduced cognitive demand in the new interfaces. Behavioral parameters from both passengers and operators were embedded in fleet‑scale simulations for a representative 10 MW hotel load. The combined interventions yield a modeled 2% reduction in hotel demand, equivalent to 0.2 MW and 1.44 GWh per year, with estimated annual fuel‑cost savings of 360,000–950,000 USD and avoided emissions of approximately 403–994 metric tonnes of CO₂ per vessel. The thesis identifies modular integration, privacy‑preserving edge inference, transparent consent, and operator oversight as key constraints for safe deployment, and argues that UX‑driven interventions at the passenger and control‑room level could be systematically combined with conventional control optimizations to deliver measurable energy and environmental benefits without degrading comfort or safety.