The Future of Lithium-Ion Battery Diagnostics: Insights from Degradation Mechanisms and Differential Curve Modeling

 

Featured Research paper: Degradation mechanisms and differential curve modeling for non-invasive diagnostics of lithium cells: An overview

De Santis, E., Pennazzi, V., Luzi, M., & Rizzi, A., Renewable and Sustainable Energy Reviews, Volume 211, April 2025

 

As the world pivots towards sustainable energy solutions, lithium-ion batteries (LIBs) have emerged as indispensable components in electric vehicles (EVs) and renewable energy systems. Their efficiency and longevity, however, are hindered by the phenomenon of battery aging — a multifaceted issue tied to the gradual decline in performance and safety. The recent paper, grounding on a project developed with Ferrari S.p.A., "Degradation mechanisms and differential curve modeling for non-invasive diagnostics of lithium cells: An overview" — published on the prestigious journal Renewable and Sustainable Energy Reviews — offers a detailed exploration of the degradation processes in LIBs, introducing innovative diagnostic methodologies and shedding light on future directions for research and industry.

Our research group at CIPARLABS is strongly committed to the development of technologies for energy sustainability. The topic of lithium-ion battery modeling is among the topics under study and development carried out by our laboratory at the "Sapienza" University  of Rome, Department of Information Engineering, Electronics and Telecommunications (DIET). 

The Challenge of Battery Aging

Lithium-ion batteries, the backbone of EVs, offer numerous advantages such as high energy density, lightweight construction, and zero emissions. However, they face significant challenges, particularly the progressive degradation of their components. Battery aging manifests as a decline in capacity, efficiency, and safety, influenced by factors such as temperature extremes, charging rates, and the depth of discharge (DOD). Addressing these issues is critical to optimizing battery performance and aligning with broader environmental goals like the UN Sustainable Development Goals (SDGs).

Battery degradation occurs in two primary forms:

• Calendar Aging: Degradation during storage, even in the absence of active use, exacerbated by conditions like high temperature and elevated state of charge (SOC).

• Cycle Aging: Degradation resulting from repetitive charging and discharging cycles.

These processes lead to two key degradation modes:

• Loss of Lithium Inventory (LLI): A reduction in the cyclable lithium ions due to side reactions.

• Loss of Active Materials (LAM): Structural damage or dissolution of electrode materials, impacting the battery’s ability to store and deliver energy effectively.

A Diagnostic Revolution: Differential Curve Modeling

The cornerstone of the paper is its focus on differential curve modeling—a non-invasive and powerful tool for diagnosing battery aging. Differential curves, specifically Incremental Capacity (IC) and Differential Voltage (DV) curves, are derived from charge/discharge data. These curves amplify subtle changes in battery behavior, revealing critical insights into degradation mechanisms.

• Incremental Capacity (IC) Curves: By plotting the change in charge against voltage, IC curves expose phase transitions in electrode materials, which are sensitive to degradation modes.

• Differential Voltage (DV) Curves: These represent voltage changes relative to charge, offering detailed insights into electrode-specific reactions and transitions.

These curves act as diagnostic fingerprints, capturing the nuanced dynamics of battery aging. For instance, shifts in IC curve peaks or DV curve valleys can be linked to specific degradation processes, enabling precise assessments of battery health.

Bridging Science and Application

The paper highlights the practical potential of differential curve analysis. In the automotive sector, this technique can be integrated into Battery Management Systems (BMS) for real-time monitoring and predictive maintenance. By identifying early signs of aging, manufacturers can optimize charging protocols, enhance safety, and extend battery lifespan. This not only reduces costs but also aligns with sustainability objectives by minimizing waste.

In the energy sector, differential curves can support the management of large-scale energy storage systems, ensuring reliability and efficiency. Policymakers, too, can leverage these insights to refine regulations and standards for EVs, accelerating the transition to sustainable transportation.

Future Directions and Innovations

While differential curve modeling offers substantial promise, challenges remain. Noise sensitivity during data processing and variability in experimental conditions necessitate standardized protocols for broader applicability. The integration of machine learning represents an exciting frontier. By training algorithms on IC/DV curve data, researchers can automate diagnostics, identify anomalous patterns, and predict battery failures with unprecedented accuracy.

The non-destructive nature of this approach makes it particularly appealing. Unlike invasive post-mortem analyses, differential curve modeling preserves battery integrity, offering a cost-effective and scalable solution for both academic and industrial applications.

Conclusion

The insights presented in the paper underscore the transformative potential of advanced diagnostic techniques for lithium-ion batteries. By unraveling the complexities of degradation mechanisms and leveraging differential curve modeling, researchers and industry leaders can pave the way for safer, more efficient, and sustainable energy storage solutions. As the global push for electrification and decarbonization accelerates, such innovations are not just timely but essential.

The road ahead is one of collaboration and innovation, bridging gaps between scientific research, industrial practices, and policy frameworks. With tools like differential curve modeling, we are better equipped to meet the challenges of the energy transition and drive a future powered by clean and reliable energy.

 

Please cite as:

• APA format:
  

De Santis, E., Pennazzi, V., Luzi, M., & Rizzi, A. (2025). Degradation mechanisms and differential curve modeling for non-invasive diagnostics of lithium cells: An overview. Renewable and Sustainable Energy Reviews, 211, 115349. https://doi.org/10.1016/j.rser.2025.115349

• BibTex format:
  

@article{ENRICO2025115349,
title = {Degradation mechanisms and differential curve modeling for non-invasive diagnostics of lithium cells: An overview},
journal = {Renewable and Sustainable Energy Reviews},
volume = {211},
pages = {115349},
year = {2025},
issn = {1364-0321},
doi = {https://doi.org/10.1016/j.rser.2025.115349},
url = {https://www.sciencedirect.com/science/article/pii/S136403212500022X},
author = {De Santis Enrico and Pennazzi Vanessa and Luzi Massimiliano and Rizzi Antonello},
keywords = {Ageing, Diagnosis, Degradation mechanisms, Degradation modes, Differential curves, Differential voltage, Lithium-ion batteries, Incremental capacity, State of health}
}

 

 

 

 

Large Concept Models: A Paradigm Shift in Language and Information Representation

Left: visualization of reasoning in an embedding space of concepts (task of summarization). Right: fundamental architecture of an Large Concept Model (LCM). ⋆: concept encoder and decoder are frozen.

Large Concept Models (LCMs) – presented by Facebook research [1] – represent a transformative leap in the field of natural language processing, aiming to address limitations in traditional Large Language Models (LLMs) by operating on a higher semantic plane. While LLMs focus on token-level processing, predicting the next word or subword in a sequence, LCMs abstract their operations to the level of "concepts." These concepts are language- and modality-agnostic, encapsulating the semantic essence of sentences or higher-order units of meaning, which positions LCMs as closer analogs to human reasoning.

Traditional LLMs have demonstrated unparalleled success across a range of tasks, from text generation to multimodal applications like image captioning or code synthesis. However, their reliance on token-level prediction imposes a fundamental limitation. Tokens capture syntactic, low-level structure but do not inherently represent the semantic relationships or higher-order dependencies humans naturally process. LCMs, by contrast, explicitly model these relationships, leveraging conceptual units that enable reasoning and planning over long sequences or across complex ideas.

In this context, LCMs align well with the broader theoretical framework of language and information representation. At the heart of this shift lies the notion of concepts as high-level semantic units. These units, often corresponding to sentences, abstract away from the specifics of language and modality, providing a universal representation of meaning. This abstraction is vital for tasks requiring multilingual and multimodal capabilities, as it decouples the reasoning process from the constraints of individual languages or data types. For example, by operating in a sentence embedding space like SONAR, which supports multiple languages and modalities, LCMs can seamlessly perform zero-shot generalization, extending their reasoning capabilities to unseen languages or domains without additional fine-tuning. This reimagining of neural networks emphasizes the hierarchical and structured nature of human cognition, where meaning is not a property of individual words but emerges from their relationships within larger contexts.

This conceptual foundation brings several advantages. First, LCMs inherently handle longer contexts more efficiently. Unlike token-based transformers, whose computational complexity scales quadratically with sequence length, LCMs process much shorter sequences of embeddings. This not only reduces resource demands but also enables more effective modeling of dependencies over extended sequences. Additionally, by decoupling reasoning from language or modality, LCMs achieve unmatched flexibility, allowing for local interactive edits, seamless multilingual integration, and the potential to extend into novel modalities like sign language. Furthermore, the explicit hierarchical structure of LCMs facilitates coherent long-form generation, making them particularly suited for tasks like summarization, story generation, or cross-lingual document alignment.

However, LCMs are not without limitations. Their reliance on pre-trained embedding spaces like SONAR introduces a dependency that may constrain performance when embeddings do not perfectly align with the model’s reasoning architecture. While the abstraction to sentence-level concepts is a significant step forward, further extending this abstraction to paragraphs or sections remains an area of ongoing research. Additionally, LCMs currently lag behind LLMs in tasks requiring fine-tuned instruction-following, such as creative writing or detailed conversational exchanges. This gap highlights the need for better integration of low-level token dynamics with high-level conceptual reasoning.

The architecture of LCMs is pivotal in achieving their ambitious goals. The process begins with an encoder, typically leveraging a pre-trained system like SONAR, which maps sentences into a fixed-dimensional semantic embedding space. These embeddings form the foundation for the LCM core – a transformer-based model that predicts the next embedding in an autoregressive manner. The core transforms and processes embeddings through layers of attention and feed-forward mechanisms, using additional components like PreNet and PostNet to normalize and denormalize embeddings to match the transformer’s internal dimensions. Variants of the core architecture explore different strategies for embedding prediction. For example, diffusion-based models introduce noise to embeddings during training and iteratively denoise them during inference, while quantized models discretize the embedding space into learned centroids for efficient prediction. Finally, the decoder converts the predicted embeddings back into text or other modalities, ensuring semantic consistency with the original input.

The ability of LCMs to generate coherent text hinges on this seamless integration of encoding, reasoning, and decoding. By operating in a structured embedding space, the model can capture the relationships between concepts and generate meaningful continuations or transformations of input sequences. The flexibility of the decoder allows for outputs in different languages or modalities, depending on the task requirements, without necessitating retraining or additional data.

Looking forward, the potential for LCMs is immense. Future improvements could involve developing embedding spaces specifically optimized for conceptual reasoning, moving beyond generic solutions like SONAR. Extending the level of abstraction to encompass paragraphs or sections could unlock new capabilities in long-form text generation and document analysis. Hybrid architectures that combine the strengths of token-level models with concept-level reasoning might also bridge the current gap in instruction-following tasks. Moreover, incorporating additional modalities, such as visual or gestural data, could further expand the applicability of LCMs to areas like education, accessibility, and human-computer interaction.

The conceptual shift represented by LCMs has profound implications for the future of AI. By modeling reasoning and meaning at a semantic level, these systems provide a path toward more human-like understanding and interaction. As research progresses, LCMs have the potential to redefine how machines represent and process information, bringing us closer to the goal of truly intelligent systems.

Large Concepts Models, therefore, are another piece to recompose the puzzle that will allow more intelligent and performing systems that can approach the so-called AGI (Artificial General Intelligence). Below is a detailed list of open research topics in the field of LLMs:

• Multimodal Integration: Training models that process diverse modalities (text, images, audio, video) to develop broader contextual understanding and reasoning abilities.
• Hierarchical Reasoning: Incorporating explicit reasoning structures, like Large Concept Models (LCMs), to operate at multiple abstraction levels for coherent, long-term planning.
• Retrieval-Augmented Generation (RAG): Combining LLMs with external knowledge bases to enhance factual accuracy and reduce hallucinations.
• Memory-Augmented Models: Developing persistent memory architectures for LLMs to store and recall knowledge dynamically across interactions.
• Embodiment and Simulation: Training AI systems in physical or simulated environments to foster embodied cognition and interactive learning.
• Self-Supervised Learning (SSL): Leveraging vast unlabeled data to improve representations without explicit supervision, advancing generalization and robustness.
• Continual Learning: Developing mechanisms for incremental knowledge acquisition without forgetting past information (mitigating catastrophic forgetting).
• Prompt Engineering and Fine-Tuning: Refining prompts or using adaptive fine-tuning strategies to align LLM outputs with specific tasks or ethical standards.
• Modular Architectures: Splitting tasks among specialized sub-models that collaborate for broader, more efficient problem-solving.
• Meta-Learning: Enabling models to learn how to learn, generalizing quickly to new tasks with minimal training.
• Neuro-Symbolic Approaches: Combining neural networks with symbolic reasoning for interpretable and logic-driven decision-making.
• Causal Reasoning: Integrating causality into LLMs to enable better reasoning and decision-making beyond statistical correlations.
• Energy-Efficient Training: Investigating low-resource training methodologies, such as sparse transformers and quantized architectures, for scalability.
• Evolutionary Algorithms: Applying optimization inspired by biological evolution to explore model architectures and strategies.
• Large Context Windows: Extending LLMs' capacity to handle longer contexts for coherent long-form reasoning and memory.
• Alignment and Alignment-Based Architectures: Ensuring LLMs align with human values and goals, incorporating reinforcement learning from human feedback (RLHF).
• Distributed Multi-Agent Systems: Creating networks of interacting agents to simulate collaborative intelligence and emergent behavior.
• Transformer Alternatives: Exploring architectures beyond transformers (e.g., RNN variants, Perceiver, or liquid neural networks) for flexibility and efficiency.
• Sparse Models: Utilizing sparsity in parameter usage to scale model sizes without corresponding resource costs.
• Open-Ended Exploration: Developing AI systems capable of self-guided exploration and intrinsic motivation to learn autonomously.
• Simulated Interiority: Training LLMs to create intermediate symbolic representations that mimic introspective thought.
• Ethical and Societal Alignment: Embedding ethical reasoning, bias mitigation, and societal impact analysis into AI development.

To learn more about Large Concepts Models:

Download the Paper

GitHub repository 

________________________

[1] The LCM team, LargeConceptModels: LanguageModelinginaSentenceRepresentationSpace, arXiv, 2024