The rapid growth of retired electric vehicle batteries presents a major opportunity for low-cost stationary storage through second-life use, but current repurposing standards rely on lengthy diagnostic tests that limit throughput. Fast and reliable screening methods are therefore needed to assess suitability for a second life before full recertification. This work investigates whether a short-duration (∼ 100 s) direct-current pulse voltage response contains sufficient information for such rapid screening. We address four questions: (1) how accurately state of health (SOH) can be inferred from a short-duration pulse sequence; (2) how pulse-derived features vary with state of charge (SOC) and SOH across chemistries; (3) whether physics-informed equivalent circuit model (ECM) parameters improve extrapolation across SOC; and (4) whether lightweight, SOC-agnostic models can provide practically useful accuracy and be deployed on embedded hardware. We develop a framework that extracts both statistical and physics-informed ECM features from the pulse response, characterizes their dependence on SOC and chemistry, and trains compact ElasticNet-based ensembles for LFP and NMC cells. Feature-quality analysis reveals when ECM parameters provide complementary diagnostic information and when statistical features alone suffice. Across both chemistries, we quantify the accuracy of SOH estimation under matched and mismatched SOC conditions, establishing the regimes in which cross-SOC generalization is feasible. Finally, we deploy and experimentally validate the LFP model on microcontroller-based hardware, demonstrating a practical pathway for rapid, low-cost second-life battery screening that reduces reliance on traditional, multi-hour tests.

Accurately forecasting lithium-ion battery capacity degradation is crucial for optimizing the second-life utilization of these batteries, enabling reliable operation, reduced maintenance costs, and extended life cycle performance. However, achieving consistent forecasting accuracy across cells and over time remains challenging due to significant cell-to-cell variability and substantial changes in real-world usage conditions during the transition from first to second life. In this study, we propose a new physics-informed machine learning method that integrates an aging-aware electrochemical model with a recurrent neural network, creating a physics-informed recurrent neural network (PI-RNN). This hybrid model leverages both physics-based insights and data-driven learning to predict capacity fade under diverse usage conditions, including transitions from first- to second-life applications. We evaluate PI-RNN using two datasets: an open-source NASA dataset comprising 28 lithium cobalt oxide/graphite cells, and a newly collected dataset of 39 commercial lithium iron phosphate/graphite cells, where cells were initially cycled to 80% capacity in their first life before undergoing milder cycling in their second life. While PI-RNN performs comparably to data-driven models in the first-life phase, it demonstrates a clear advantage in second-life forecasting, reducing root mean squared error by approximately 40%–70% compared to baseline models when forecasting periods span the transition from first to second life, even when trained on as few as two cells. Parametric studies highlight the advantages of incorporating physics-based modeling, and uncertainty quantification ensures the reliability of long-term capacity forecasting. In addition, we conducted benchmarking studies to systematically assess the advantages and limitations of the proposed model, thus identifying the scenarios where this approach excels.

Early initial impoundment can generate additional revenue but bring more flood risk in late-stage construction diversion. In view of the possible flood risk and catastrophic consequences caused by high dam failures induced by early impoundment, a comprehensive assessment is proposed. Taking the Lianghekou high rockfill dam on the Yalong River, southwest China, as an example, this study established the late-stage diversion risk model and predicted the failure probabilities for the original, 15 days ahead, and 30 days ahead schemes varied with the initial impoundment time using the Monte Carlo method. Then, considering overtopping-induced gradual breaking of rockfill dams, the NWS dam-break flood forecasting model (DAMBRK) was used to estimate the break development and the outflow hydrograph. Due to no significant differences being found in the outflow hydrographs of the three schemes, life loss was used an index for the consequences of inundation. Combining the failure probability, life loss, and early impoundment revenues brought by earlier power generation, a satisfied initial impoundment scheme was acquired using the multi-objective decision model. The results revealed this method can find a reasonable initial impoundment time in view of the late-stage diversion risk assessment.