Asoba Asoba

Real-World Use Cases

See how Asoba delivers measurable business value across revenue optimization, risk quantification, and uptime maximization.

Sibaya Casino

Limited Historical Data: Achieving Forecasting Accuracy in 6 Months

Industry Commercial
Capacity 1.5 MW
Location Durban, South Africa

Challenge

Only 6 months of historical data available for a 1.5MW solar plant. Industry standard requires 24+ months for accurate forecasting models, creating a 27-40 month deployment timeline before reliable predictions are possible.

7%
SMAPE Accuracy
6
Months to Accuracy

Results

  • Achieved 7% SMAPE accuracy in 6 months using transfer learning
  • Leveraged neural networks trained on similar Durban and Johannesburg sites (12-24 months data)
  • Global LSTM modeling approach used external sites as training set
  • Forecasting accuracy from day one vs. industry standard 27-40 month deployment

Cummins Portfolio

Overcoming 65% Data Loss: Maintaining Operations Through Infrastructure Failure

Industry Industrial
Portfolio Size Multi-MW
Data Loss 65% Missing

Challenge

Severe data loss—65% missing data points across multi-MW portfolio. Sensor and telemetry failures made traditional monitoring impossible. Most systems treat missing data as downtime for analytics.

100%
Data Reconstruction
96%
Faster Detection

Results

  • 100% data reconstruction using ARIMA-based interpolation + ML ensemble
  • Zero grid violations maintained despite infrastructure failure
  • 96% faster fault detection despite 65% data loss
  • Maintenance logic continued without interruption—system treated missing data as signal, not downtime

Avaron Infrastructure

Multi-Stakeholder Coordination: Unified Intelligence Across Owner, O&M, and Insurer

Industry IPP/Asset Management
Stakeholders Owner, O&M, Insurer
Sites Multiple

Challenge

Coordinating Owner, O&M contractor, and Insurer across multiple sites. Fragmented systems required manual reconciliation. Each stakeholder needed different intelligence from the same operational data.

70%
Penalty Reduction
100%
Data Reconciliation

Results

  • Unified data layer eliminated reconciliation between stakeholders
  • Economic optimization now considers insurance impact of maintenance timing
  • Multi-stakeholder coordination without data conflicts
  • 70% reduction in unplanned grid code penalties through risk analytics

Sustainable Energy Africa

EnergyAnalyst v2: Improving Renewable Energy Interconnection Process Across South African Municipalities

Pilot scheduled for Q1-Q2 2026 with 3 municipalities in partnership with University of Johannesburg

Industry Municipal/Regulatory
Platform apply.sseg.org.za
Utilities 50-60 Active
Documents 3,759 Regulatory
Pilot Partner University of Johannesburg
Pilot Timeline Q1-Q2 2026

Challenge

South African municipalities must evaluate thousands of rooftop solar and embedded generation applications every year, each governed by a unique mix of local regulations and national legislation. Sustainable Energy Africa (SEA) operates apply.sseg.org.za, a nationwide portal that digitizes the application workflow for 50–60 utilities. However, applicants preparing their interconnection applications must manually interpret lengthy PDF policies from multiple sources to understand requirements, compliance rules, and approval processes for their specific municipality.

Solution: EnergyAnalyst v2 is a retrieval-augmented generation (RAG) assistant built for applicants to use when preparing their applications. It processes 3,759 regulatory documents from municipalities, national government, NERSA, and other regulatory bodies. Applicants can ask questions in natural language and receive instant, municipality-specific answers with citations in under 8 seconds, eliminating the need to manually search through PDFs. This dramatically accelerates application preparation and reduces errors from misunderstanding regulations.

91%
Positive Feedback
<8s
Median Latency
4,200
Initial Test Queries
3,759
Regulatory Documents

Initial Testing Results

  • EnergyAnalyst v2 is a retrieval-augmented generation (RAG) assistant that delivers instant, municipality-specific answers with citations
  • Sub-eight-second median latency across 4,200 initial test queries
  • 91% positive feedback from initial testing
  • Continued pre-training (Stage 2A) and instruction-tuning (Stage 2B) pipeline processes 3,759 regulatory documents
  • Data governance model co-designed with SEA balances public transparency with policy data confidentiality

Upcoming Pilot

  • Three-municipality pilot scheduled for Q1-Q2 2026 in partnership with University of Johannesburg
  • Evaluation harness will be deployed during pilot for systematic assessment
  • Architecture designed for scaling to all 167 South African utilities

Edge Deployment Architecture

Real-Time Intelligence: 2-5 Second Forecasts on ARM64 Edge Devices

Platform ARM64 Edge
Devices Raspberry Pi CM4, Orange Pi 5, Rock 5B
Power 3-5W Active

Challenge

Cloud lag costs money in trading. Need real-time decisions without cloud dependency. Connectivity loss shouldn't stop operations. Traditional cloud-based systems introduce latency and single points of failure.

2-5s
Forecast Generation
3-5W
Active Power

Results

  • 2-5 second forecast generation on ARM64 edge devices
  • Operates independently during connectivity loss—forecasting continues at edge
  • 3-5 watts active, 1 watt idle power consumption
  • Edge-native computing enables energy trading, VPPs, and distributed infrastructure optimization

Ready to See These Results for Your Portfolio?

Start with a technical assessment or explore our APIs to see how Asoba can transform your energy operations.

Request Demo Explore APIs

Get Help & Stay Updated