Feature Engineering

Shifting time-series data backward to capture the sequential dependency of gas generation. Lag features convert a static snapshot of DGA values into a dynamic sequence, allowing models to learn the rate of change.

• Example: A lag-24 feature for acetylene (C₂H₂) represents the gas concentration 24 hours prior.
• Mechanism: Enables tree-based models and LSTMs to detect trends that a single reading cannot reveal.
• Key Insight: Rapidly rising hydrogen (H₂) with a lagged baseline is a critical indicator of partial discharge.

Calculating aggregate metrics over a moving temporal window to smooth sensor noise and highlight persistent shifts. Rolling statistics transform volatile raw telemetry into stable, trend-indicating features.

• Rolling Mean: Dampens transient spikes in moisture content to reveal the true drying or wetting trend of solid insulation.
• Rolling Standard Deviation: Quantifies the instability of a load tap changer's motor current, indicating mechanical wear.
• Rolling Min/Max: Captures the daily peak hot-spot temperature, which directly dictates the rate of cellulose aging per IEEE C57.91.

Engineering composite features by calculating the ratios of specific dissolved gases to classify fault types. Gas ratios are the foundational input for the Duval Triangle and IEC 60599 interpretation methods.

• Duval Ratio 1 (CH₄/H₂): Differentiates thermal faults from electrical discharges based on the relative production of methane versus hydrogen.
• Duval Ratio 2 (C₂H₂/C₂H₄): High values indicate high-energy arcing due to acetylene generation.
• CO₂/CO Ratio: A critical indicator of paper degradation; a low ratio suggests pyrolysis of cellulose insulation.

Computing the first-order difference between consecutive time steps to capture the instantaneous velocity of change. This feature type is highly sensitive to incipient faults that manifest as sudden gas spurts.

• Delta Acetylene: A non-zero delta is an immediate red flag for active arcing, as C₂H₂ is not produced by normal aging.
• Rate of Rise: Calculating the ppm-per-day increase in combustible gases to trigger alarms before absolute thresholds are breached.
• Application: Essential for online DGA monitors where trending velocity is more actionable than static concentration.

Transforming time-domain signals into the frequency domain to extract periodic patterns invisible in raw waveforms. Spectral features are critical for analyzing vibration and acoustic partial discharge data.

• Fast Fourier Transform (FFT): Decomposes vibration signals from load tap changers to identify specific mechanical resonance frequencies indicating misalignment.
• Power Spectral Density: Quantifies the energy distribution of acoustic emissions to distinguish internal partial discharge from external corona.
• Cepstral Analysis: Used to detect periodic echoes in Frequency Response Analysis (FRA) traces, revealing winding deformation.

Creating new features by multiplying, dividing, or combining two or more independent variables to capture non-linear physical relationships. Interaction features encode domain expertise directly into the model.

• Load × Ambient Temperature: A combined thermal stress index that better predicts hot-spot temperature than either variable alone.
• Moisture × Oxygen: Models the synergistic catalytic effect that drastically accelerates cellulose depolymerization.
• Gas Ratio × Load: Contextualizes a high acetylene ratio with the transformer's operating state to suppress false positives during transient heavy loading.

Create temporal predictors by shifting time-series data backward. For Dissolved Gas Analysis (DGA) , generating lag features allows models to learn the rate of gas generation rather than just absolute concentrations.

• Example: Shifting acetylene (C₂H₂) readings by 1, 6, and 24 hours to capture the velocity of arcing faults.
• Key Concept: Transforms a static snapshot into a dynamic trajectory, enabling LSTM and Temporal Fusion Transformer models to detect acceleration trends.

Compute moving averages, standard deviations, and min/max windows over sensor streams to smooth noise and highlight persistent shifts. Essential for Hot-Spot Temperature monitoring.

• Rolling Mean: A 24-hour moving average of winding temperature filters out transient load spikes.
• Rolling Std Dev: Detects increasing vibration variability in Load Tap Changer (LTC) motor signatures before mechanical failure.
• Exponentially Weighted Moving Average (EWMA) : Applies higher weight to recent observations for drift detection.

Construct domain-specific diagnostic ratios from raw DGA ppm values to classify fault types. These ratios form the basis of the Duval Triangle and IEC 60599 interpretation standards.

• Key Ratios: CH₄/H₂ (methane/hydrogen), C₂H₂/C₂H₄ (acetylene/ethylene), C₂H₄/C₂H₆ (ethylene/ethane).
• Duval Triangle 1: Uses %CH₄, %C₂H₄, and %C₂H₂ to map faults to thermal fault, partial discharge, or arcing zones.
• Rogers Ratio: A four-ratio coding scheme for classifying thermal and electrical faults per IEEE C57.104.

Separate raw sensor signals into trend, seasonal, and residual components to isolate anomalous behavior from normal cyclic patterns. Critical for Online DGA Monitors with daily load cycles.

• Trend Component: Long-term insulation degradation trajectory extracted from Degree of Polymerization (DP) proxies.
• Seasonal Component: Daily and weekly load-driven temperature cycles removed to expose true thermal anomalies.
• Residual Analysis: The remaining noise component where incipient fault signatures first appear before trending algorithms detect them.

Engineer cross-variable features that capture the physical coupling between transformer subsystems. Physics-Informed Neural Networks (PINNs) leverage these to enforce thermodynamic constraints.

• Load × Ambient Temperature: Interaction term modeling the non-linear relationship governing Hot-Spot Temperature rise per IEEE C57.91.
• Moisture Content × Temperature: Combined feature predicting bubble evolution temperature and dielectric strength reduction.
• Gas Generation Rate Ratios: Cross-gas rate comparisons (e.g., CO₂/CO generation ratio) indicating paper versus oil degradation.

Transform time-series vibration and acoustic signals into the frequency domain using Fast Fourier Transform (FFT) to extract spectral signatures of mechanical faults.

• Spectral Centroid: Tracks shifts in dominant vibration frequency indicating LTC contact wear.
• Harmonic Distortion: Odd-harmonic amplitudes in current waveforms signaling core saturation or geomagnetic induced currents.
• Wavelet Packet Energy: Multi-resolution decomposition capturing transient Partial Discharge pulses across frequency bands for source localization.