lithium-ion battery, deep learning, lifespan prediction, fast charging strategy optimization, state of health estimation costs.1 In practical applications, a precise understanding of SOH and RUL not only improves battery efficiency but also effectively reduces failure rates and extends service life.2 Conventional battery lifespan prediction methods largely depend on equivalent circuit models and physical modeling. However, these techniques involve complex modeling processes and are often inadequate for adapting to complex and dynamic real-world conditions. With the rapid progress in machine learning and deep learning, datadriven methods for predicting the RUL of lithium-ion batteries have attracted growing interest.3 These approaches are more flexible and eliminate the need for detailed physical modeling, allowing predictions to be made based solely on historical battery operation data. Deep learning models, such as recurrent neural networks (RNNs), offer particular advantages in processing time-series data due to their inherent memory capabilities, making them highly suitable for various lifespan prediction tasks. Meanwhile, fast-charging technology—key to improving user experience—faces the challenge of balancing charging speed with battery lifespan. Optimization strategies for lithium battery fast charging vary across applications and often require integrating multiple charging policies. For instance, both electric vehicles and mobile phones must consider both charging speed and battery longevity.4 As fast-charging technology evolves and scenarios grow more complex, multi-objective optimization has become necessary. To mitigate issues such as severe internal polarization caused by high-current charging—which adversely affects performance and longevity—researchers have proposed various optimization methods that simultaneously consider charging time and battery degradation. Therefore, research on predicting the lifespan of lithium-ion batteries and optimizing fast-charging strategies using deep learning has significant theoretical and practical value. It contributes to improved battery efficiency, extended service life, and reduced energy consumption and environmental pollution. Research objectives and content This research was aimed at enhancing the accuracy of estimating the state-of-health (SOH) of lithium-ion batteries and the reliability of predicting the remaining useful life (RUL) by means of deep-learning techniques and at the same time optimizing fast-charging strategies to strike a balance between charging efficiency and battery life because lithium-ion batteries had been widely adopted in new-energy vehicles and energy-storage systems and accurate SOH estimation and RUL prediction were of great importance for system safety and cutting down maintenance costs. Particular research aims encompassed constructing a deep-learning-based state-of-health (SOH) evaluation model for lithium-ion batteries, carrying out early-remaining useful life technology, and formulating fast-charging optimization strategies that equalize charging time and battery lifespan, and in contrast to traditional approaches founded on equivalent circuits and physical models, data-driven deep-learning methods are more pliable as they removed the necessity for intricate physical modeling and could be modeled solely with historical battery operational data, thus effectively prognosticating the battery’s remaining useful life (RUL). This research mainly involves the following areas: First, probing into the health traits of lithium-ion batteries and the early-remaining useful life techniques with the application of deep learning by making use of the data from the early cycles (about 10% of the battery’s entire life cycle) to forecast the battery’s full life cycle5; second, examining the merits of deep-learning models, like recurrent neural networks, in dealing with battery time-series data and building an effective lifespan prediction model; third, putting forward an optimization approach that strikes a balance between the charging speed and the battery life, aiming at the control goals of the lithium-battery fast-charging optimization strategies in diverse application scenarios; finally, experimentally verifying the efficacy of the proposed model, offering theoretical support and technical guidance for the lithium-ion battery health management systems. Literature review Basic principles of lithium-ion batteries As a high-performance secondary battery, lithium-ion batteries mainly function on the basis of the back-and-forth movement of lithium ions between the positive and negative electrodes and the fundamental structure of a battery consists of crucial components like the positive electrode, negative electrode, electrolyte, separator, and battery shell, the positive electrode material typically employs lithium compounds, for instance, lithium iron phosphate (LiFePO4) or lithium cobalt oxide (LiCoO2) and negative electrode materials are mostly carbon-based materials, particularly graphite, whose layered structure contains a great number of micropores, which is conducive to the insertion of lithium ions.6 During the charging process, lithium ions in the positive electrode substance were deintercalated and passed through the separator by means of the electrolyte and finally got embedded in the graphite structure of the negative electrode and in order to keep the internal charge neutrality of the battery, an equal quantity of electrons had to flow from the positive electrode to the negative electrode via an external circuit and it was notable that electrons couldn’t directly go through the separator and had to be conducted through an external circuit.7 When it came to discharge, lithium ions were deintercalated from the negative electrode and then reintercalated into the positive electrode substance, releasing electrons to supply energy8 and this intercalation-deintercalation process was highly reversible, which ensured the cycle life and operational safety of the battery.9 The performance of lithium-ion batteries is influenced by a variety of factors including internal physicochemical reactions such as the properties of electrode materials and the composition of electrolytes and external operating conditions like ambient temperature, charge/discharge rate, and depth of discharge,10 lithium salts within the electrolyte have a remarkably significant influence on battery performance directly affecting the efficiency of lithium-ion transportation,11,12 as the application of lithium-ion batteries has been continuously expanding, accurate estimation of their state of health and real-time prediction of their remaining useful life have become more and more important which was crucial for ensuring the safe operation of battery systems and reducing maintenance costs. Related technological advances The domain of lithium-ion battery technology has witnessed substantial advancement in recent years, especially in terms of lifespan prediction and health condition evaluation, and as machine learning and deep learning developed rapidly, data-driven techniques for forecasting the remaining useful lifespan of lithium-ion batteries had attracted extensive attention, and in contrast to traditional research approaches based on equivalent circuits and physical models, data-driven prediction methods were more adaptable and did not need complicated physical modeling, and modeling could be carried out merely with historical data from the battery’s operation, thus effectively predicting the battery’s remaining useful lifespan. Recurrent neural networks, thanks to their distinctive memory architecture, possess substantial benefits when it comes to handling time-series data and are extensively employed in diverse lifespan prediction tasks. Liang et al. put forward a method for predicting the lifespan of lithium batteries based on multi-scale decomposition and deep learning. 13 Moreover, Zhao et al. proposed a method based on deep learning for estimating the remaining useful life span of lithium-ion batteries which offered new research perspectives in this area.14 In the analysis of lithium-battery failures, the use of deep-learning technology allows for the extraction, prediction, and diagnosis of the characteristics related to lithium-battery failures, which effectively enhanced the efficiency of lithium batteries, decreased the failure rates, and prolonged their service lives, and as deep-learning technology continuously progressed and became more refined, the field of lithium-battery failure analysis was expected to attain more breakthroughs and innovations, but it should be noted that the lifespan of lithium batteries in actual operating conditions was influenced by multiple factors including internal elements (such as the consistency of battery clusters, the design of battery modules, and internal physicochemical alterations) and external ones (such as the operating temperature, the charge rate, and the state of charge),15 which made high-precision prediction of the realworld lifespan difficult and had spurred the development of more high-precision lifespan prediction technologies. Overview of deep learning models Basics of deep learning Overview of main algorithms Deep learning technology was increasingly utilized in lithium-ion battery research, core algorithms consisted of recurrent neural networks, convolutional neural networks, and autoencoders, and each of these algorithms had its own distinct features and was appropriate for different kinds of battery data analysis and prediction tasks. Recurrent neural networks (RNNs), along with their variants like long short-term memory (LSTM) networks, are good at handling time-series data because of their distinctive memory architecture, which makes them highly suitable for predicting battery life, and LSTMs could capture long-term relationships within battery cycles,