Asset Intelligence

Overview

The Asset Intelligence Hub is Energy Copilot’s most sophisticated DNSP module. It combines an asset register with ML-powered failure prediction, condition-based health scoring, and a risk-consequence matrix — giving asset managers a single pane of glass for strategic expenditure decisions.

The XGBoost failure prediction model achieves AUC 0.961 and 92.3% accuracy on the test set. See Asset Failure Prediction ML for full model documentation.

Asset Registry

The asset registry (gold.dnsp_asset_register) maintains a comprehensive record of all distribution network assets:

Asset Type	Key Attributes
Transformers	MVA capacity, voltage, age, insulation type, last test date, oil condition
Switchgear	Type (SF6, oil, air), voltage class, last test, mechanism condition
Underground cables	Length, insulation material, age, fault history, thermal load
Overhead conductors	Length, conductor type, age, sag, span rating, vegetation clearance
Poles	Material (wood, steel, concrete), age, tip load, soil classification
Protection relays	Manufacturer, model, firmware version, last test, trip history

Asset Health Scoring

Each asset receives a Health Index (HI) score from 0 (end of life) to 100 (new):

HI = Σ (weight_i × condition_factor_i)

Condition factors include:

• Age factor: normalised by expected service life
• Load factor: operating load vs thermal rating
• Fault history factor: number and severity of past faults
• Maintenance recency: days since last inspection or test
• Observed defects: known issues from last inspection

Health Index bands:

Score	Band	Recommended Action
80–100	Good	Routine maintenance
60–79	Satisfactory	Enhanced monitoring
40–59	Poor	Prioritise for maintenance/replacement
20–39	Very Poor	Urgent replacement program
0–19	Critical	Immediate action required

Failure Prediction Model

The XGBoost failure prediction model runs as a Model Serving endpoint and scores each asset monthly:

# Feature set used by the model
features = [
    'age_years',
    'health_index',
    'fault_count_5yr',
    'peak_load_ratio',        # peak load / thermal rating
    'days_since_maintenance',
    'insulation_condition_score',
    'weather_exposure_index'  # based on geographic zone and climate data
]

Model outputs:

• failure_probability_12m: probability of failure in the next 12 months (0–1)
• failure_risk_class: High / Medium / Low
• top_risk_factors: SHAP-derived explanations of the top 3 contributors

Screenshot: Asset Intelligence Hub showing risk matrix heat map with assets plotted by failure probability (x-axis) vs consequence (y-axis).

Risk-Consequence Matrix

Assets are plotted on a 5×5 risk matrix:

             ┌───────────────────────────────────────────────────┐
         5   │ Med  │ High │ High │ Crit │ Crit │
  C      4   │ Low  │ Med  │ High │ High │ Crit │
  O      3   │ Low  │ Low  │ Med  │ High │ High │
  N      2   │ Neg  │ Low  │ Low  │ Med  │ High │
  S      1   │ Neg  │ Neg  │ Low  │ Low  │ Med  │
  Q          └───────────────────────────────────────────────────┘
  U               1      2      3      4      5
  E                   FAILURE PROBABILITY
  N
  C
  E

Consequence dimensions assessed:

• SAIDI impact: expected customer-minutes interrupted if this asset fails
• Network criticality: radial vs meshed supply
• Customer type: hospitals, high-dependency customers, CBD
• Replacement lead time: how long to restore if a major failure occurs

Expenditure Justification

AER requires DNSPs to justify capital expenditure proposals under the “capex test” — expenditure must be prudent, efficient, and consistent with a prudent regulatory obligation.

The Asset Intelligence Hub generates expenditure justification packs:

1. Asset condition evidence: health scores, condition inspection reports, defect photos
2. Failure risk quantification: failure probability × consequence × cost-of-failure
3. Alternatives analysis: repair vs replace vs augment
4. Cost-benefit analysis: NPV of different options over the regulatory period
5. Benchmark comparison: cost per unit vs AER benchmarks

Cross-System Integration

The Asset Intelligence Hub integrates data from multiple DNSP systems:

System	Data Provided	Integration Method
EAM/SAP PM	Asset register, maintenance history	API or daily export
SCADA/DMS	Load data, fault events	API or real-time feed
GIS	Geographic attributes, span lengths	Shapefile import
Customer system	Customer connections by feeder	API
Vegetation management	Last clearance date, clearance status	API

API Endpoints

# Asset health summary
GET /api/dnsp/assets/summary?dnsp=ausgrid

# Individual asset details
GET /api/dnsp/assets/{asset_id}

# High-risk assets (failure probability > threshold)
GET /api/dnsp/assets/high-risk?dnsp=ergon&threshold=0.3&horizon_months=12

# Risk matrix data
GET /api/dnsp/assets/risk-matrix?dnsp=energex

# Expenditure justification pack
POST /api/dnsp/assets/capex-justification
{
  "asset_ids": ["TXF-001-234", "TXF-001-235"],
  "project_type": "replacement",
  "cost_aud": 450000
}

# Failure prediction scores (latest)
GET /api/dnsp/assets/failure-prediction?dnsp=essential_energy&min_probability=0.2

Tip

The AI Copilot get_asset_health_scores tool retrieves health scores and failure predictions. Ask: “Which transformers in Zone 3 are at high failure risk in the next 12 months and what is the estimated SAIDI impact if they fail?”