Search

This work presents a novel approach to the challenge of battery charging under real-world constraints, related to uncertainties in system parameters and unmeasurable internal states of batteries. By leveraging the imitation learning paradigm, this study introduces an innovative solution to address the inherent challenges associated with traditional predictive control strategies. A key contribution of this work is the successful application and adaptation of the Dataset Aggregation (DAGGER) algorithm to an 'agnostic scenario', characterized by uncertain battery parameters and unobservable internal states. Furthermore, this work is, to the authors' best knowledge, the first attempt to amalgamate deep predictive control within the imitation learning framework, offering a fresh perspective and broadening the array of possible solutions to the difficulties in battery charging. Results derived from a realistic battery simulator implementing an electrochemical model demonstrate marked enhancements in battery charging performance, particularly in satisfying temperature constraints. The performance of the proposed algorithm surpasses that of existing approaches, including a benchmark behavioral cloning method based on supervised learning. These advancements highlight the potential of the imitation learning paradigm in tackling complex control problems in battery management systems.

The growing complexity and interdisciplinary nature of Materials Science research demand efficient data management and exchange through structured knowledge representation. Domain-Level Ontologies (DLOs) for Materials Science have emerged as a valuable tool for describing materials properties, processes, and structures, enabling effective data integration, interoperability, and knowledge discovery. However, the harmonization of DLOs, and, more generally, the establishment of fully interoperable multi-level ecosystems, remains a challenge due to various factors, including the diverse landscape of existing ontologies. This work provides, for the first time in literature, a comprehensive overview of the state-of-the-art of DLOs for Materials Science, reviewing more than 40 DLOs and highlighting their main features and purposes. Furthermore, an alignment methodology including both manual and automated steps, making use of Top-Level Ontologies' (TLO) capability of promoting interoperability, and revolving around the engineering of FAIR standalone entities acting as minimal data pipelines ('bridge concepts'), is presented. A proof of concept is also provided. The primary aspiration of this undertaking is to make a meaningful contribution towards the establishment of a unified ontology framework for Materials Science, facilitating more effective data integration and fostering interoperability across Materials Science subdomains.

This work investigates the effectiveness of deep neural networks within the realm of battery charging. This is done by introducing an innovative control methodology that not only ensures safety and optimizes the charging current, but also substantially reduces the computational complexity with respect to traditional model-based approaches. In addition to their high computational costs, model-based approaches are also hindered by their need to accurately know the model parameters and the internal states of the battery, which are typically unmeasurable in a realistic scenario. In this regard, the deep learning-based methodology described in this work was been applied for the first time to the best of the authors' knowledge, to scenarios where the battery's internal states cannot be measured and an estimate of the battery's parameters is unavailable. The reported results from the statistical validation of such a methodology underline the efficacy of this approach in approximating the optimal charging policy.

In the last decade, the techniques of news aggregation and summarization have been increasingly gaining relevance for providing users on the web with condensed and unbiased information. Indeed, the recent development of successful machine learning algorithms, such as those based on the transformers architecture, have made it possible to create effective tools for capturing and elaborating news from the Internet. In this regard, this work proposes, for the first time in the literature to the best of the authors' knowledge, a methodology for the application of such techniques in news related to cryptocurrencies and the blockchain, whose quick reading can be deemed as extremely useful to operators in the financial sector. Specifically, cutting-edge solutions in the field of natural language processing were employed to cluster news by topic and summarize the corresponding articles published by different newspapers. The results achieved on 22,282 news articles show the effectiveness of the proposed methodology in most of the cases, with 86.8% of the examined summaries being considered as coherent and 95.7% of the corresponding articles correctly aggregated. This methodology was implemented in a freely accessible web application.

This work presents the MarketPlace Agent and Expert Ontology (MAEO), an ontology for modeling experts, expertise, and more broadly, knowledge providers and knowledge seekers for the subject areas of Materials Modeling. MAEO had its inception within the "MarketPlace" European project, whose purpose is to bring about a single entry point for gathering scientific and industrial stakeholders in Materials Modeling. As such, this project aimed to build an online platform where experts and knowledge providers can be searched, found and brought into contact with users, or knowledge seekers, and with one another. MAEO was developed in order to fulfill the requirements of this online platform and thus support it, but is also part of a wider ecosystem of Materials Modeling-related ontologies, at whose core lies the Elementary Multiperspective Material Ontology (EMMO). MAEO is thus an EMMO-compliant application ontology, and has been loosely aligned with a number of existing ontologies, including Friend-Of-A-Friend (FOAF) and five recently-developed EMMO-based domain ontologies for the classification of materials, models, manufacturing processes, characterization methods and software products related to Materials Modeling. Here, a detailed description of the axiomatization of MAEO and its interconnected ontologies is provided, along with results coming from its deployment and experimentation in a StarDog triplestore.

Discussion forums are popular tools in Massive Open Online Courses (MOOCs), used by students to express feelings, exchange ideas, and ask for help. Unfortunately, the huge number of enrolled students hinders educational scaffolding activities, including the active participation of instructors in discussion forums. Therefore, students seeking to clarify the concepts learned may not receive the answers they need, reducing engagement and promoting dropout. This work presents a methodology for supporting learners within discussion forums, by analyzing conversations among students and providing them with recommendations in terms of relevant learning resources. The methodology involves several steps: the initial definition of an ontology that details the topics of the course, the real-time analysis of student posts within the discussion forums to extract different attributes including intent of the post, the concepts it is about, the sentiment, and the level of urgency and confusion. The extracted information is then used by a rules-based mechanism to assess whether the learner needs a recommendation. If so, the system suggests the most suitable learning resources. The paper also includes an initial evaluation of the proposed methodology.

The fast charging of a lithium-ion battery is a complex task, which needs to be addressed by a proper control methodology to find the highest charging current while guaranteeing safety. Among the different approaches, model predictive control appears particularly suitable due to its ability in dealing with nonlinear systems and constraints. However, its use in a realistic scenario is limited due to the high computational burden required by the online solution of an optimal control problem. To overcome this issue, we consider a neural network-based algorithm, which can reduce the online computational effort by approximating the solution of the model predictive control. Such a deep learning-based approach is here applied for the first time to the real-time management of a lithium-ion cell, described by an electrochemical model with thermal dynamics. The results highlight the effectiveness of the proposed methodology in terms of computational burden reduction.

One of the most crucial challenges faced by the Li-ion battery community concerns the search for the minimum time charging without damaging the cells. This goal can be achieved by solving a large-scale constrained optimal control problem, which relies on accurate electrochemical models. However, these models are limited by their high computational cost, as well as identifiability and observability issues. As an alternative, simple output-feedback algorithms can be employed, but their performance strictly depends on trial and error tuning. Moreover, particular techniques have to be adopted to handle safety constraints. With the aim of overcoming these limitations, we propose an optimal-charging procedure based on deep reinforcement learning. In particular, we focus on a policy gradient method to cope with continuous sets of states and actions. First, we assume full state measurements from the Doyle-Fuller-Newman (DFN) model, which is projected to a lower dimensional feature space via the principal component analysis. Subsequently, this assumption is removed, and only output measurements are considered as the agent observations. Finally, we show the adaptability of the proposed policy to changes in the environment's parameters. The results are compared with other methodologies presented in the literature, such as the reference governor and the proportional-integral-derivative approach.

Batteries are complex systems that need to be properly managed to guarantee safe and optimal operations. Model predictive control (MPC) is an advanced control strategy that, thanks to its characteristics, can be embedded into battery management systems (BMS) to derive optimal charging strategies. However, deterministic MPC, which relies on a nominal model only, is not adequate in a realistic scenario in which cells parameters are not known exactly. In this paper, stochastic MPC is proposed for the optimal charging of a Li-ion battery pack to account for the presence of parameter uncertainties. The adopted scheme relies on the polynomial chaos expansion paradigm for the propagation of uncertainties through the model equations and allows to satisfy safety constraints with a guaranteed probability. The results highlight the advantages of stochastic MPC over different scenarios when compared to a deterministic MPC approach.