thinkwithyili Battery Cell Datasheet Extraction Skill
Overview
This skill extracts and standardizes battery cell specifications from supplier datasheets (PDFs). It handles variations in terminology, units, and table formats across different manufacturers.
Section A: Data Identification
1. Locate Specification Tables
Primary locations: "Electrical Characteristics", "Physical Specifications", "General Specifications", "Nominal Specification" tables.
Secondary sources: Header/footer areas, performance graphs, notes sections.
2. Table Layout Patterns
| Pattern | Identifier | Handling |
|---|---|---|
| A: Key-Value | 2-column (param | value) | Direct mapping |
| B: Multi-Condition | 3+ columns with merged cells | Inherit parent from previous row for empty cells |
| C: Visual Grouping | Indentation-based, no borders | Use text position for hierarchy |
| D: Multi-Product | Multiple model columns | Extract each column as separate cell object |
3. Common Field Variations
| Standard Field | Variations |
|---|---|
| Nominal Capacity | Rated Capacity, Typical Capacity, Cell Capacity |
| Nominal Voltage | Cell Voltage, Operating Voltage |
| DCIR | DC Internal Resistance, AC Impedance, Internal Resistance |
| Max Discharge | Maximum Continuous Discharge, Discharge Current |
Section B: Extraction Procedures
1. Basic Cell Information
Extract: Manufacturer, Model Number, Chemistry (NMC/LFP/NCA/NCMA), Cell Format (cylindrical/pouch/prismatic), Format Code (18650/21700/4680).
2. Electrical Properties
| Property | Notes | Conversion |
|---|---|---|
| Capacity | Look for Nominal/Rated/Typical; may have Typ/Min values | mAh → Ah (÷1000) |
| Voltage | Nominal, Max (charge cutoff), Min (discharge cutoff) | — |
| Impedance | ACIR at 1kHz most common; record test conditions | Ω → mΩ (×1000) |
| Current | Continuous discharge, charge, pulse (with duration) | C-rate → A: C × Capacity(Ah) |
3. Temperature Properties
Parse formats: "0~45°C", "0°C to 45°C", "-20 to 60 deg C" → {min_c, max_c}
Extract separately for: charge, discharge, storage ranges.
4. Physical Properties
| Format | Dimensions |
|---|---|
| Cylindrical | Diameter (mm), Height (mm) |
| Pouch | Length × Width × Thickness (mm) |
| Prismatic | Length × Width × Height (mm) |
Weight: grams (convert kg → g if needed).
5. Cycle Life
Parse: "500 cycles at 80% DOD" → {cycles: 500, dod: 80}
Record: temperature, C-rates, EOL SOH threshold.
Section C: Normalization Rules
1. Standard Field Names
Use canonical names: capacity.nominal_ah, voltage.nominal_v, impedance.acir.value_mohm, current.max_continuous_discharge_a, temperature.operating.charge.min_c.
2. Unit Conversions
| From | To | Operation |
|---|---|---|
| mAh | Ah | ÷ 1000 |
| Ω | mΩ | × 1000 |
| kg | g | × 1000 |
| C-rate | A | C × Capacity(Ah) |
3. Missing/Ambiguous Data
- Not found: set
null, add tofields_missing - Ambiguous: extract best estimate, flag "medium" confidence
- Multiple values: extract all (typ/min/max)
- Store
unit_originalfield
4. Reference Files
normalization_rules.yaml: Supplier-specific patternsvalidation_schema.json: JSON schema for output validationscripts/extractor.py: Deterministic parsing functions
Section D: Validation
1. Required Fields
Manufacturer, Model Number, Nominal Capacity, Nominal Voltage, Cell Format.
2. Value Ranges
| Property | Valid Range |
|---|---|
| Capacity | 0.5 - 200 Ah |
| Nominal Voltage | 2.5 - 4.5 V |
| Max Voltage | ≤ 5.0 V |
| DCIR | 1 - 200 mΩ |
| Weight | 10 - 5000 g |
3. Confidence Scoring
- High: All fields found, clear values, known format
- Medium: Some inferred, ambiguous conditions
- Low: Many missing, unfamiliar format, OCR issues
Section E: Output Format
1. JSON Structure
Output follows validation_schema.json. Top-level objects:
cell_info: manufacturer, model, chemistry, format, confidencemechanical: dimensions, weight, volumeelectrical: capacity, voltage, impedance, currenttemperature: operating ranges (charge/discharge/storage)lifetime: cycle life specificationsderived: calculated propertiesextraction_metadata: fields_extracted, fields_missing, warnings
2. Derived Properties (calculate if not provided)
| Property | Calculation |
|---|---|
| Energy (Wh) | Voltage × Capacity |
| Max Power (W) | Voltage × Max Discharge Current |
| Gravimetric Energy (Wh/kg) | Energy ÷ Weight |
| Volumetric Energy (Wh/L) | Energy ÷ Volume |
3. Export to Sheets
python sheets_exporter.py output.json --sheet "Battery Cell Database"
Worked Example: Samsung INR21700-50E
Input (Pattern B - multi-condition):
2.1 Capacity: Typical 5000mAh, Minimum 4850mAh
2.2 Voltage: Nominal 3.6V, Charge 4.2V, Cutoff 2.5V
2.3 Impedance: AC 1kHz ≤18mΩ (50% SOC, 25°C)
Extraction:
- Pattern B identified → parse nested structure
- Capacity: typical=5.0Ah, minimum=4.85Ah
- Voltage: nominal=3.6V, max=4.2V, min=2.5V
- Impedance: ACIR=18mΩ at 1kHz, 50% SOC, 25°C
Output (abbreviated):
{
"cell_info": {"manufacturer": "Samsung", "model_number": "INR21700-50E", "chemistry": "NMC", "cell_format": "cylindrical"},
"electrical": {
"capacity": {"nominal_ah": 5.0, "minimum_ah": 4.85},
"voltage": {"nominal_v": 3.6, "max_v": 4.2, "min_v": 2.5},
"impedance": {"acir": {"value_mohm": 18.0, "frequency_hz": 1000, "temperature_c": 25, "soc_percent": 50}}
},
"derived": {"energy_wh": {"value": 18.0, "calculation": "3.6V × 5.0Ah"}}
}
Quick Tips
- Start with manufacturer ID → predicts format patterns
- Check notes/footnotes → contain critical conditions
- Cross-reference values → verify energy = V×Ah
- When uncertain → flag with warnings, don't guess