Scientific Objectives
Common BEMS solutions offer cloud-based services which adapt to consumption patterns (thermostat, light, etc.) according to a set of specific rules (supervised or unsupervised learning) that usually only consider energy-related data (generation and consumption). However, in many cases, users finally do not follow their recommendations since they are not adapted to other requirements, such as comfort or preferences. In addition, cloud-based solutions have a demanding data transmission bandwidth and an increased cyberattack surface. In this context, the AI-BEMS project has been built on the basis of the following general hypothesis:
Optimized energy use in buildings and better adherence to the recommendations of BEMS can be achieved by considering subjective factors, such as users’ comfort and preferences. Regarding optimization, explainable artificial intelligence models and social computing approaches can improve the recommended energy efficiency measures. Regarding adherence, distributed approaches can increase the adoption of these systems since centric approaches are seen as a bottleneck in security and privacy.
This general hypothesis is in turn subdivided into the following specific hypotheses:
H1
It is possible to design a BEMS that takes advantage of the benefits of the Edge Computing paradigm to optimize energy consumption in buildings and homes.
H2
Considering users’ comfort and preferences as well as the reputation of the EE measures can increase the consumers’ adherence to the recommendations of the BEMS, fostering long-term energy-efficient habits.
H3
Explainable Artificial Intelligence (XAI) algorithms can identify the factors that affect energy consumption and recommend appropriate efficiency measures.
H4
Edge Computing coupled with Distributed Ledger Approaches (DLTs) can increase the security and privacy of the users’ energy consumption data.
Following these hypotheses, the main objective of the AI-BEMS project is:
To optimize the energy consumption at buildings/homes and enable the demand response by researching and building an Edge-Computing architecture based on virtual organizations and distributed Explainable Artificial Intelligence (XAI) algorithms. Virtual organizations will dynamically optimize the consumption patterns and recommend energy efficiency (EE) measures will not only increase energy savings but will also maximize the residents’ comfort. The virtual organization will be able of autonomously learning from users.
The general objective has been subdivided into the following specific objectives together with the expected key results (KR), each one corresponds to a specific hypothesis:
O1
To design and build a scalable Edge Computing architecture for BEMS, capable of ingesting data from the IoT devices deployed in buildings/homes and from any other source of heterogeneous data. Moreover, the architecture will run artificial intelligence algorithms locally.
O2
To investigate the ability of virtual organizations based on social computing, to recommend energy efficiency measures, taking into account the users’ comfort and preferences, and the reputation of the measures among users.
O3
To investigate new neuro-symbolic XAI algorithms for energy efficiency that will further reduce energy consumption and improve the interpretability of the models.
O4
To develop a DLT framework (e.g., blockchain-based) for the edge-computing architecture that allows for the secure identification and data exchange of IoT devices, guaranteeing the privacy of the users’ data.