Claim-Level Evidence for Physics-Informed Machine Learning in Resilient Microgrids

Evidence Dataset This folder contains structured extraction tables for a systematic evidence review on physics-informed, data-driven, safe, and constrained learning methods for microgrid and power-system operation. The dataset separates source-level decisions from claim-level evidence. A source is counted once in source_inventory.csv and eligibility_decisions.csv. A claim is a method-task-context unit extracted from an included source and is counted in evidence_claims.csv. Core Tables File Purpose source_inventory.csv Bibliographic and topical inventory for all reviewed sources. eligibility_decisions.csv Included, contextual, and excluded decisions with rationale. evidence_claims.csv Main claim-level evidence table for included sources. denominators.csv Separate source-level and claim-level counts. coding_rules.csv Rules used for eligibility, claim splitting, evidence class, fidelity tier, and auditability tier. claim_unit_examples.csv Examples showing how sources were split into claim units. classification_decisions.csv Difficult classification cases and final decisions. Supporting Tables File Purpose retrieval_log.csv Retrieval and screening metadata available from the local corpus. citation_screening.csv Bounded citation-following decisions. coding_resolution.csv Available coding-resolution information. review_context.csv Comparison with review sources in the corpus. evidence_cross_tabulations.csv Cross-tabulated claim counts. evidence_summaries.csv Distribution summaries for included claim evidence. runtime_complexity.csv Latency, runtime, and computational-complexity evidence. safety_assumptions.csv Safety and protection assumptions for operational claims. uncertainty_evidence.csv Uncertainty, calibration, and risk-related evidence. scalability_evidence.csv Large-system, topology, and scaling evidence. failure_modes.csv Feasibility conditions and reported failure modes. benchmark_requirements.csv Evidence-derived benchmark requirements. evidence_coverage.csv Mapping from review needs to supporting tables. dataset_index.csv Compact file index. Reading Order Start with source_inventory.csv to identify all reviewed sources and citation keys. Use eligibility_decisions.csv to separate included, contextual, and excluded sources. Use evidence_claims.csv for the main evidence synthesis. Join it to eligibility_decisions.csv using study_id or to source_inventory.csv using citation_key. Use denominators.csv when reporting counts. Do not mix source-level and claim-level denominators. Use coding_rules.csv and claim_unit_examples.csv to interpret classification decisions and claim splitting. Key Coding Values final_status uses Included, Contextual, and Excluded. validation_fidelity_tier uses: F0: static, steady-state, operating-point, analytical, or offline feasibility evaluation. F1: ordinary time-domain, RMS, DAE, averaged, simulation-level, or non-real-time dynamic evaluation. F2: high-fidelity transient, EMT, switching-level, or detailed device/protection simulation. F3: real-time simulation, hardware-in-the-loop, laboratory, field, or prototype evaluation. auditability_tier uses: A0: operational context or limits not clearly stated. A1: operational context stated with at least one metric. A2: explicit test condition plus constraint, violation, stability, or feasibility outcome. A3: A2 plus robustness breadth, feasible runtime or latency, and baseline or ablation. NR means the information was not reported in the available corpus.