Abstract
Utilities in South Africa require auditable, standards-aligned methods to verify metering accuracy and to surface non-technical losses (NTL). Routine audits are fragmented across tools, and diagnostic findings are difficult to trace back to regulatory clauses.
To develop and validate a standards-traceable MATLAB framework that unifies clause-level compliance checking, key performance indicator (KPI) computation, and theft-detection analytics for interval meter data, and to evaluate its performance on diverse field sites. The study implemented a modular pipeline that (i) parses heterogeneous CSV exports into a canonical schema; (ii) executes executable predicates mapped to NRS 048/057 and IEC 62053 clauses; (iii) computes KPIs (monthly completeness, PF distributions, 15/30/60-min maximum demand); and (iv) applies three anomaly detectors flatline (zero-variance), reverse-power, and daily-drop. Five anonymised points of delivery (S1–S5; Elster A1700, 3- and 4-wire, various CT/VT ratios) were analysed. Outputs were compared with domain-expert review; agreement was summarised using Cohen’s κ. All sites met or exceeded the 95% completeness benchmark and achieved “Compliant” outcomes for applicable accuracy clauses. KPI profiles showed stable maximum-demand behaviour and PF patterns consistent with industrial/municipal feeders; sites with sustained PF<0.90 suggested under-compensated inductive loading rather than instrumentation error. The theft module flagged flatline and daily-drop episodes across several sites and isolated reverse-power risk where polarity issues were suspected. Expert agreement with automated flags was high (κ=0.78–0.92). Most positive detections were attributable to telemetry stalls, short communication outages, or polarity anomalies; persistent, deliberate tamper was not evidenced in the study window. The framework produced reproducible tables, shaded overlays, and a PDF report linking each result to its predicate and clause. Translating standards into executable predicates, coupled with integrated KPI and anomaly analytics, enables transparent, auditable metering assurance. The framework supports both regulatory compliance and revenue-protection triage by turning raw interval data into defensible evidence. Limitations include reliance on expert adjudication for ground truth, rule sensitivity to data quality, and limited LV coverage. Future work will extend to smart-meter cohorts, on-line dashboards, and hybrid rule/ML models for adaptive thresholds and automated prioritisation.