Sibaya Casino Case Study · May 2026

From Billing Confusion to Inverter-Level Clarity

Sibaya Casino installed a 2.31 MW solar system to reduce grid costs. But municipal bills didn't drop as expected. Asoba's intelligence platform diagnosed a R1M annual grid-bill leak, tracing 92.6% of the shortfall to chronic inverter-level hardware failures. By closing this "Savings Gap," the site restores R624k in annual net profit.

Site: Durban, SA 2.31 MW capacity R624k Net Savings Gap 5 Visualizations
I

The Site Profile

Sibaya Casino and Entertainment Kingdom, operated by Sun International, is a premier hospitality venue in Durban consuming approximately 15.4 million kWh annually. To offset 20% of its grid dependency, the site commissioned a 2.31 MW rooftop solar plant featuring seven Huawei 330 kW inverters. For a 24/7 operation, every unit of solar energy is a direct R2.50 credit against the municipal bill.

1. Site Topology: Sibaya Solar Infrastructure
MUNICIPAL GRID METER SIBAYA LOAD CENTER 15.4M kWh/yr 2.31 MW SOLAR ARRAY 7x Huawei 330kW Inverters POI
Site Topology: Solar generation (2.31 MW) connects at the Point of Interconnection (POI) to offset site load. Telemetry is pulled from 7 individual inverters.
II

The Problem: The Variance Gap

In April 2026, Sibaya's engineering team noticed a troubling pattern: while Huawei FusionSolar reported healthy generation, municipal bills remained stubbornly high. Theoretically, 2.9M kWh of solar should have dropped the grid bill significantly. Instead, the site observed a 16.1% variance over theoretical expectations in 2025.

2. The Variance Gap: Theoretical vs Actual Grid Consumption (2025)
THEORETICAL GRID 12.4M kWh ACTUAL MUNICIPAL 14.4M kWh +16.1% VARIANCE 2.0M kWh Overspend
The Billing Gap: Municipal meter readings were 16.1% higher than theoretical projections, suggesting that solar offsets were not being fully realized.
III

The Diagnosis: Physics-Based Verification

The first task was to confirm that the billing discrepancy was real and not an artefact of meter error or data misalignment. Asoba reconciled Eskom meter readings against solar production telemetry across the full analysis period (October 2025 – May 2026).

METER RECONCILIATION: OCTOBER 2025 – MAY 2026
ENERGY BALANCE ITEM VALUE (kWh)
Total Site Consumption 7,024,260
Solar Production (Metered) 1,113,190
Grid Purchases 5,911,070
RECONCILIATION ERROR 0.00%

The reconciliation was clean: Consumption = Solar + Grid, with zero discrepancy. This immediately ruled out meter fraud, billing errors, or data integrity issues. It confirmed that the issue was not the bill—it was the physics. The solar system was present and reporting as "online," but it was simply not producing as much as it should have been.

Using IEC 61724 methodology, Asoba disaggregated the shortfall. 92.6% of the production gap was traced to chronic hardware-related faults, while only 7.4% was weather-attributed. The system was, in Asoba's terminology, "chronically underperforming while running."

74.7% Actual Performance Ratio (Target: 80%)
92.6% Losses due to Hardware Failures
116k Detected Fault Events
IV

Inverter Deep Dive: Identifying the Culprits

Asoba disaggregated the analysis to the individual inverter level. INV-190 was the primary culprit, operating at a critical 53.6% PR and costing the site R79,099 in excess grid purchases during the analysis period alone. The top four underperformers accounted for 63% of the total financial impact.

3. Per-Inverter Performance Ratio (PR) vs 80% Target
80% TARGET INV-190 53.6% INV-192 66.9% INV-191 69.1% INV-193 74.3% INV-195 82.3% INV-194 82.9% INV-196 84.9%
Inverter Performance: 4 of the 7 inverters were operating significantly below the 80% Performance Ratio target. INV-190 was operating in a CRITICAL state at 53.6%.
V

The Predictive Edge: JEPA Anomaly Detection

Asoba's JEPA (Joint-Embedding Predictive Architecture) world model learns "normal" operation for every inverter. Unlike rule-based systems, JEPA detects subtle degradation patterns. At Sibaya, the model demonstrated an AUROC score of up to 0.98, catching all critical faults while maintaining a minimal false positive rate.

4. Detection Performance: JEPA Model AUROC Scores
INV-192 0.98 INV-195 0.92 INV-191 0.89 INV-196 0.83 INV-190 0.69* AUROC Score (1.0 = Perfect Detection)
Predictive Performance: The JEPA model achieved near-perfect anomaly detection (0.98 AUROC) on key units, significantly outperforming traditional rule-based alerts. (*INV-190 limited by sparse data).
VI

Outcome: Closing the Savings Gap

The diagnosis is clear. For every kWh lost to hardware failure, Sibaya pays a R2.50 grid premium. While the system was reporting as "online," it was leaking value. By deploying targeted maintenance to the primary culprits (INV-190 to 193), Sibaya closes a R1.02M annualized gap in its grid bill, resulting in R624k in annual net savings returned to the bottom line.

5. Outcome Dashboard: Realizing the Economic Value
ANNUAL GRID BILL LEAKAGE R1.02M Cost of excess grid purchases (@ R2.50/kWh) 410,556 kWh ANNUAL SHORTFALL NET SAVINGS RECOVERY R624k Annual net profit returned to bottom line BASED ON R1.52 NET BENEFIT/kWh
Economic Impact: The "Savings Gap" represents the difference between the R2.50/kWh municipal avoided cost and the underlying cost of solar. By eliminating invisible inverter faults, Sibaya stops a R1.02M annual overspend on grid electricity.
RECOMMENDATIONS
Immediate
Physical Site Inspection Deploy technicians to INV-190 through 193. Inspect AC/DC connections and check for water ingress or physical damage.
Short-Term
Performance Audit Commission an independent audit against the 80% PR target. Review inverter firmware and string-level performance.
Strategic
Predictive Alerting Activate Asoba JEPA-based real-time alerting to intercept faults at onset and eliminate billing variance.

Understand the gap in your portfolio.

Upload your asset data. Run a physics-based performance analysis. See the revenue gap that SCADA cannot detect. Understand the settlement precision that SAWEM demands. See what the Utility Ledger category can deliver to your portfolio.

SCHEDULE AN ANALYSIS