Efficient energy management relies on uncovering meaningful consumption patterns from large-scale electricity load demand profiles. With the widespread adoption of sensor technologies such as smart meters and IoT-based monitoring systems, granular and real-time electricity usage data have become available, enabling deeper insights into consumption behaviors. Clustering is a widely used technique for this purpose, but previous studies have primarily focused on a limited set of algorithms, often treating clustering as a black-box approach without addressing interpretability. This study explores a wide number of clustering algorithms by comparing hard clustering algorithms (K-Means, K-Medoids) versus soft clustering techniques (Fuzzy C-Means, Gaussian Mixture Models) in segmenting electricity consumption profiles. The clustering performance is evaluated using five different clustering validation indices (CVIs), assessing intra-cluster cohesion and inter-cluster separation. The results show that soft clustering methods effectively capture inter-cluster characteristics, leading to better cluster separation, whereas intra-cluster characteristics exhibit similar behavior across all clustering approaches. This study assesses which CVIs provide reliable evaluations independent of clustering algorithm sensitivity. It provides a comprehensive analysis of the different CVIs’ responses to changes in data characteristics, highlighting which indices remain robust and which are more susceptible to variations in cluster structures. Beyond evaluating clustering effectiveness, this study enhances interpretability by introducing two decision tree models, axis-aligned and sparse oblique decision trees, to generate human-readable rules for cluster assignments. While the axis-aligned tree provides a complete explanation of all clusters, the sparse oblique tree offers simpler, more interpretable rules, emphasizing a trade-off between full interpretability and rule complexity. This structured evaluation provides a framework for balancing transparency and complexity in clustering explanations, offering valuable insights for utility providers, policymakers, and researchers aiming to optimize both clustering performance and explainability in sensor-driven energy demand analysis.

Interactive learning plays a key role in extending the occupant behavior implementation towards smart buildings. Efficient feedbacks can be obtained from the end user by involving occupants and increasing their awareness about energy systems.Working in highly energy-efficient buildings can be a great opportunity, but users need to feel empowered. This means making them aware of the building features and allowing them to manage some of the appliances. In this way, disorientation or annoyance is avoided and people feel more in control. This paper proposes a solution to interact with occupants to estimate the number of occupants. A novel way of supervised learning is analyzed to estimate the occupancy in a room where actual occupancy is interactively requested to occupants when it is the most relevant to limit the number of interactions. Occupancy estimation algorithm relies on machine learning: it uses information gathered from occupants with the measurements collected from common sensors, for instance, motion detection, power consumption, and CO2 concentration. Two different classifiers are investigated for occupancy estimation with interactions: a decision tree C4.5 and a parameterized rule-based classifier. In this paper, the question of when interacting with occupants is investigated. This approach avoids the usage of a camera to determine the actual occupancy. A complete real-time interaction enviornment has been developed and is used to estimate the occupancy in an office case study. The graphical user interface has been designed to carry out a real-time experiment.

This article proposes a novel deep clustering model based on the variational autoencoder (VAE), named GamMM-VAE, which can learn latent representations of training data for clustering in an unsupervised manner. Most existing VAE-based deep clustering methods use the Gaussian mixture model (GMM) as a prior on the latent space. We employ a more flexible asymmetric Gamma mixture model to achieve higher quality embeddings of the data latent space. Second, since the Gamma is defined for strictly positive variables, in order to exploit the reparameterization trick of VAE, we propose a transformation method from Gaussian distribution to Gamma distribution. This method can also be considered a Gamma distribution reparameterization trick, allows gradients to be backpropagated through the sampling process in the VAE. Finally, we derive the evidence lower bound (ELBO) based on the Gamma mixture model in an effective way for the stochastic gradient variational Bayesian (SGVB) estimator to optimize the proposed model. ELBO, a variational inference objective, ensures the maximization of the approximation of the posterior distribution, while SGVB is a method used to perform efficient inference and learning in VAEs. We validate the effectiveness of our model through quantitative comparisons with other state-of-the-art deep clustering models on six benchmark datasets. Moreover, due to the generative nature of VAEs, the proposed model can generate highly realistic samples of specific classes without supervised information.

A general approach is proposed to determine occupant behavior (occupancy and activity) in offices and residential buildings in order to use these estimates for improved energy management. Occupant behavior is modelled with a Bayesian network in an unsupervised manner. This algorithm makes use of domain knowledge gathered via questionnaires and recorded sensor data for motion detection, power, and hot water consumption as well as indoor CO₂ concentration. Different case studies have been investigated with diversity according to their context (available sensors, occupancy or activity feedback, complexity of the environment, etc.). Furthermore, experiments integrating occupancy estimation and hot water production control show that energy efficiency can be increased by roughly 5% over known optimal control techniques and more than 25% over rule-based control while maintaining the same occupant comfort.