A significant educational and citizen-science action is conducted among project works, including organising workshops, conferences, and lectures defining the PRELUDE T9.5 – Educational Event Series.
The first event of this series was a conference/lecture on smart building issues, discussing “Proactive, forecasting, and optimisation approaches for intelligent buildings”. Held on the 13th of January 2023, the event has been followed in a hybrid mode, both physically at POLITO and remotely online. This conference included six technical presentations focussing on PRELUDE topics and an introduction to the educational event series led by Giacomo Chiesa by POLITO.
- Michal Pomianowski (Aalborg University, Denmark) – PRELUDE introduction, aims and challenges
The PRELUDE project – Prescient building Operation utilising Real-Time data for Energy Dynamic Optimization – aims at supporting advanced innovative, smart and low-cost solutions supporting intelligent buildings and proactive optimisation services. Among the others, the project pursues the minimisation of energy uses, the maximisation of self-consumption and renewable sources balance, and the improvement of comfort conditions supporting the maximisation of free-running building potentials. This presentation focuses on the overall PRELUDE project presentation, its aims and challenges. The project concept is developed of a modular PRELUDE platform built and integrated around the middleware solution FusiX. Several of the PRELUDE modules basis on historical data available to analyse but also to be proactive and forecast: weather, energy, and comfort.
- Kais Dai (TREELOGIC, Spain) – Scaling-up technologies for multiple models’ aggregation
In the context of the PRELUDE project, TREE is proposing solutions to efficiently aggregate models in multiple buildings scenarios (i.e., neighbourhood, district, city) towards generating knowledge and intelligence at scale based on the aggregation of the individual ones (i.e., taking the predictive models and optimization features to the neighbourhood/district/city scale). Several techniques have been proposed to achieve this goal, such as consumption profiles clustering, ensemble models and federated learning. All these models perform 24-hour-ahead forecasting of the district heating energy consumption.
- Nikos Sofias (CORE IC, Greece) – Machine Learning solutions for weather and energy forecasting in energy-efficient buildings
In the scope of the PRELUDE project, CORE-IC is proposing data-driven methodologies, (Machine Learning) for local climate Weather and Energy Forecasting. This process includes energy-efficient buildings as pilots in different European climate zones. Forecasting successfully the Energy consumption – production plays a crucial role in energy management, planning, and functionality optimization of residential buildings. The context of the presentation is fundamental data analysis and time-series forecasting tools. Results from PRELUDE project real demo cases, about 24h ahead of weather and energy forecasts, have been presented in this lecture.
- Christian Heschl (Forschung Burgenland, Austria) – The energy-saving potential of data-driven predictive control
Decarbonization of the energy system requires higher energy efficiency and increased demand-side flexibility. Intelligent building energy management systems utilizing model predictive control (MPC) can be crucial in achieving this goal by optimizing energy demand and integrating more renewables through load shifting. However, traditional MPC implementation can be complex and resource-intensive. Data-driven predictive control (DPC) offers a promising alternative, as it can significantly improve energy savings while simplifying the implementation process. The lecture presents the principle structure of a DPC solution and its potential for energy savings.
- Giacomo Chiesa, Paolo Grasso (Politecnico di Torino, Italy) – New platform for building energy white box modelling and 24h forecasting
Inside the PRELUDE project, new functionalities and usage scenarios have been added to the PREDYCE tool, a python library acting as a dynamic simulation platform connecting energyplus with monitored data. PREDYCE supports sensitivity analyses and massive simulation studies, directly comparing KPIs calculated from simulation and real-time monitored data. A new usage scenario to optimise shading and ventilative cooling activation schedules for the next day, considering forecasted weather conditions and building characteristics, is presented here. According to building smartness levels, results are conceived to be sent to mechanical actuators via BMS or to the end-users for self-actuation. The presentation shows the 24h forecasting architecture and discusses sample results on a PRELUDE demo flat simulated under typical weather. Comparisons between optimized conditions and standard shading/ventilation controlling action are given to describe the approach better.
- Mattia Repossi, Lorenzo Farina (STAM, Italy) – The EnPower platform
The EnPower platform supports users in controlling and optimizing HVAC, adopting a holistic approach that keeps into account Thermal Comfort and IAQ in real time with constant synchronization with the connected home devices and is integrated within a modular, collaborative platform. This is achieved by the load distribution optimization that considers economic, environmental, technical and user-related data to forecast the optimal schedule for the times ahead. Therefore, the typical starting point is the optimal exploitation of the contractual conditions, particularly with reference to the canonical tariff bands. This is associated with the smoothing of the power peaks, giving the user the possibility to decide whether to request a reduction in the contractual power, with the consequent economic advantages, in the face of imposing a technical limit on the quantity of loads electricity that can be powered at the same time and therefore to the reduction of comfort. Homogeneous distribution of the withdrawal demand allows easier user management by the electricity provider, reducing network problems such as overloads and overvoltages.
An intense debate followed the presentations discussing the replicability of the illustrated approaches, current limitations and future paths in the intelligent building proactive sector.