{"id":"b571f3e7-0011-47fa-a26d-fed97e833261","arxiv_id":"2411.08025","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A method reconstructs quasi-open-circuit voltage curves from home storage field data and uses them to estimate degradation modes, finding loss of lithium inventory dominant in the evaluated systems.","lead":"This paper reconstructs battery open circuit voltage curves from eight years of home storage field data using low-dynamic overnight discharge phases. It then applies incremental capacity and differential voltage analysis to identify loss of lithium inventory as the dominant aging mode.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"LLI-dominant conclusion is not independently validated: qOCV curves inherit SOC from the same pipeline used for validation, and the mean-shift alignment cannot remove a common SOC bias that grows with aging.","rationale":"The reader's weakest assumption correctly identifies the SOC-estimation dependence in Section 2.2.3, and I agree that unquantified SOC errors threaten the FOI-DM correlations. My stress-test goes slightly further: the problem is not only per-phase random offsets, which the mean shift can absorb, but also a common, aging-correlated bias that the alignment cannot remove. Because the only quantitative validation (field capacity tests) comes from the same Figgener et al. pipeline that supplies the SOC values, the validation is circular for the purpose of supporting the OCV reconstruction. This is a genuine soft spot, but it does not make the central claim obviously false; the method is plausible and the LMO/NMC and NMC results are internally consistent. The conditional verdict is therefore appropriate: accept only after independent SOC re-estimation or synthetic-data benchmarking. No change to the reader's verdict is needed, but the condition should explicitly include an independent SOC-axis check, not just publication of code.","tokens_in":14881,"tokens_out":4030,"duration_ms":45419,"concrete_test":"Re-analyze the six SmallLMO systems using an independent SOC estimate built from the raw current/voltage series in [2], e.g., Coulomb counting with rest-period recalibration or a fresh-cell OCV lookup, and rerun the Section 2.2-3 qOCV reconstruction and FOI tracking. If FOI1-FOI3 shifts and intensity changes, and the LLI-dominant conclusion, are preserved, the concern is resolved; if the FOI trends weaken or reverse, the central claim is an artifact of the inherited SOC axis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that loss of lithium inventory is the dominant degradation mode depends entirely on the reconstructed qOCV curves being faithful to the true OCV. The SOC axis of every partial curve comes from Figgener et al. [1], and the horizontal alignment in Section 2.2.3 only corrects offsets between individual phases, not a bias common to all phases in a period. If the [1] capacity/SOC estimator systematically under- or overestimates capacity fade, the qOCV curve will be horizontally shifted with age. Such a shift changes ICA peak intensities (since dQ/dV is computed along an SOC axis scaled by the estimated capacity) and DVA peak distances, in a way that mimics the classic LLI signatures described in [37]. The paper's stated validation is the field capacity tests from [1,2], but these are the same underlying measurements used to build the SOC axis, and the text even argues the reconstructed qOCV is 'more reliable' than those tests. There is, therefore, no independent anchor separating true OCV evolution from SOC-estimation artifacts. The LFP section explicitly concedes that SOC inaccuracies limit the analysis, which shows the mechanism is real for at least one technology. Without a re-estimation or synthetic benchmark, the dominant-LLI result for LMO/NMC and NMC systems is plausibly explained by a correlated SOC bias rather than by electrochemical degradation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a method for reconstructing quasi-open-circuit-voltage (qOCV) curves from multi-year operational field data of 21 home storage systems. Low-dynamic discharge phases (e.g., overnight household supply) are filtered, corrected for ohmic drops using a DCR look-up table, and combined into annual qOCV curves. Incremental capacity analysis (ICA) and differential voltage analysis (DVA) are applied, with literature-derived features of interest (FOIs) tracked over time and correlated with state-of-health (SOH) estimates from the authors' prior work. The central claim is that loss of lithium inventory (LLI) is the dominant degradation mode for LMO/NMC and NMC systems, with LFP less conclusive. The method is validated against field capacity tests from the same prior dataset.","tokens_in":15159,"tokens_out":2880,"duration_ms":32492,"significance":"If the central claim holds, the paper would demonstrate a valuable capability: continuous electrochemical diagnostics of home storage batteries from ordinary operational data, without laboratory OCV measurements. The dataset is substantial (106 system-years, 14 billion datapoints) and publicly available, and the method is presented with enough detail to be reproduced. The use of ICA/DVA with literature-based FOI tracking is well motivated, and the multi-chemistry application (LMO/NMC, NMC, LFP) adds breadth. However, the conclusion that LLI dominates is not supported by an independent ground truth: the SOC axis of every reconstructed qOCV curve is inherited from the same prior capacity-estimation pipeline that is later used for validation, and the horizontal alignment step cannot correct a systematic SOC bias common to all phases. The paper's own LFP section concedes that SOC inaccuracies limit the analysis for at least one chemistry. These issues do not invalidate the method but require additional validation or sensitivity analysis before the degradation-mode conclusion can be considered robust.","major_comments":[{"comment":"The reconstructed qOCV curves' SOC axis is taken entirely from the capacity/SOC estimates of Figgener et al. [1], and the alignment step in Section 2.2.3 shifts each partial curve horizontally to the mean of all phases in a period. This mean-shift cannot remove a systematic SOC error that is common to all phases in a period, such as a bias in the estimated capacity fade that grows with age. Because ICA and DVA features are computed along this SOC axis, a proportional error in the estimated capacity would shift IC peak positions and change peak intensities in ways that mimic the classic LLI signatures described in [37]. The paper reports in Section 3.2 and Figure 9 that LFP systems indeed suffer from SOC inaccuracies, confirming that this mechanism is real in the data. The authors should provide a sensitivity analysis—for example, artificially scaling or shifting the SOC axis by plausible error bounds and showing how the FOI trends and the LLI conclusion change—or anchor the SOC axis to an independent measurement on at least a subset of systems.","section":"Section 3.3"},{"comment":"The validation in Section 3.3 uses field capacity tests from [1,2], which are the same underlying measurements and the same prior estimation pipeline that supplies the SOC axis for the reconstructed qOCV curves. The statement that the qOCV curve 'is considered more reliable' than the field capacity tests further emphasizes that the validation is not independent. Since the degradation-mode inference is built on correlations between FOI shifts and the SOH estimates from [1], any systematic error in those SOH estimates propagates directly into both the qOCV shape and the correlation analysis. To break this circularity, the authors should compare at least one reconstructed qOCV curve to a laboratory-measured OCV curve from the same cell type, or generate synthetic field-like data with known degradation modes and verify that the FOI-DM correlations recover the true modes.","section":"Section 3.3"},{"comment":"The attribution of FOI trends to specific degradation modes is qualitative and non-unique. For example, FOI 2's intensity decline is stated to be 'attributable mainly to LLI, but could also include LAMNE and LAMPE,' and FOI 4's decrease is said to indicate LAMNE/LAMPE but is also correlated with SOH. The conclusion that LLI is dominant is supported by multiple FOIs, but the paper never quantifies the relative contributions of LLI, LAMNE, and LAMPE. The correlation coefficients in Table 1 show that FOI shifts correlate with SOH, but correlation does not identify the degradation mode, and the p-values are reported without multiple-comparison correction. The authors should either apply a quantitative degradation-mode decomposition (e.g., fitting half-cell OCP-based models as in [8]) or explicitly state that the LLI-dominant conclusion is a hypothesis consistent with the FOI trends rather than a quantified result.","section":"Section 3.1"}],"minor_comments":[{"comment":"The dynamics filter threshold is given as '10 % of the C-rate'; please clarify how the C-rate is defined for each system, since the systems have different power and energy ratings and the C-rate may vary over time.","section":"Section 2.2.2"},{"comment":"In the sentence 'Sometimes, SOC errors occur, leading to a horizontal deferral of single identified phases,' the word 'deferral' should likely be 'offset' or 'shift'.","section":"Section 2.2.3"},{"comment":"The normalization of FOI values in Figure 5d (intensity normalized by highest peak, position by voltage range, distance by SOC range) is described only in the caption; please define these normalizations in the main text of Section 3.1 so that the reported slopes in percent per year are unambiguous.","section":"Figure 5"},{"comment":"The p-values are reported only as '< 0.05' or '0.85'; please report exact values or at least a meaningful precision, and consider whether a multiple-comparison correction across the ten FOI/SOH correlations is appropriate.","section":"Table 1"},{"comment":"The abstract emphasizes 'eight years' of field data, but the method produces yearly qOCV curves and the analysis in Section 3.1 shows a seven-year span for the exemplary SmallLMO system; please clarify how many systems reach the full eight years and whether 'year one' refers to the first full year of operation.","section":"Abstract"},{"comment":"The abbreviations EOC and EOD are used without definition; please define them at first use (end-of-charge and end-of-discharge).","section":"Section 3.1"}],"recommendation":"major_revision","confidential_remarks":"The skeptical concern about SOC-axis circularity is well grounded in the manuscript: the qOCV reconstruction and the validation both rely on the authors' prior capacity/SOC estimates, and the paper's own LFP results concede SOC sensitivity. The manuscript's central claim (LLI dominance) is plausible but not yet independently anchored. A major revision that adds a laboratory OCV comparison or a synthetic benchmark, plus a sensitivity study of SOC bias, would address the main correctness risk. The paper is otherwise well structured and the dataset is a valuable contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Honest take: this is a useful paper with a real validation gap. The new thing is the application: reconstructing quasi-OCV curves from low-dynamic overnight discharge phases in multi-year home storage field data, then running ICA/DVA and tracking literature-defined features of interest. That hasn't been done on real HSS field data before, and the dataset is substantial (21 systems, 106 system-years). The method description is concrete – DCR look-up tables, dynamics/throughput filters, Gaussian smoothing – and the authors are upfront that LFP results are inconclusive due to the flat voltage curve. The correlation analysis with SOH is a nice touch.\n\nThe soft spot is the one the stress-test flags, and it lands. The SOC axis of every reconstructed qOCV curve comes from the same capacity estimation pipeline (Figgener et al. [1]) that provides the field capacity tests used for validation. The horizontal alignment in Section 2.2.3 only corrects offsets between individual partial curves, not a common SOC bias that grows with age. If that pipeline under- or overestimates capacity fade, the qOCV curve shifts horizontally with time, and those shifts produce exactly the ICA peak intensity changes and DVA peak distance shrinkages that the paper attributes to LLI. The paper actually makes this worse by arguing the reconstructed qOCV is 'more reliable' than the capacity tests, which removes the only independent anchor. There is no comparison to a laboratory OCV measurement, no synthetic benchmark with injected SOC errors, and no error bars on any of the reconstructed curves or FOI tracks. For LMO/NMC and NMC, the LLI-dominant conclusion is plausible but not independently established.\n\nThat said, this is not a fatal flaw for the paper's main useful contribution. The method demonstrably produces smooth, repeatable qOCV curves from messy field data, and the qualitative FOI trends are consistent with known aging signatures. As a monitoring and diagnostic tool, it is valuable. What's missing is a quantitative DM confirmation. A serious referee should push for an independent anchor – e.g., take a retired/decommissioned HSS, measure OCV in the lab, or run a synthetic data experiment where the true SOC is known and inject errors to bound the effect.\n\nI'd cite this if I were working on field-data battery diagnostics, and I think it deserves peer review despite the validation gap. Reading group: maybe.","headline":"Genuinely new field-data qOCV reconstruction and ICA/DVA degradation-mode analysis, but the LLI-dominant headline lacks an independent SOC anchor and should be treated as qualitative until validated.","tokens_in":15719,"tokens_out":2270,"would_cite":true,"duration_ms":35343,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-12T21:59:43.020077+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}