SOC & SOH ANALYTICS ｜ SOC & SOH Analytics

Make state-of-charge information usable,
and make health changes explainable

Combine high-quality data, physical constraints, and reference validation so every SOC, SOH, SOE, and SOP result has a clear model source, confidence interval, reference evidence, and version context—supporting safe operation and asset-value decisions.

The estimate is not an isolated number, but rather a controlled state of “measurement, modeling, validation, confidence and equipment identification”, which is the common denominator.

SOC estimate (median)

Confidence interval (95 per cent)

Reference events (correction points)

Validate Reference Points

SOC 54.7%

SOH 92.6%

SOE 2.34 MWh

SOP +23.6 kW

Confidence (95%) 95.2%

Authentication error (RMSE) 1.82%

System: BESS-01 Box number: CAB-03 Collection time: 2025-05-20 10:30:00 Run mode: Charge Ambient temperature: 28.6°C

01 Enter the chain of evidence

& Electricity Charge current and direction

& Voltage Pack / Cluster / Modular Pressure

Temperature Core / Model temperature distribution

Coherence Low pressure differential, internal resistance and dispersibility

REST

Still period Still long and stable

HIS

Charge History History Energy and Cycle Record

CFG

Configuration and Pop Module / cabinet / BMS Parameter

CLK

Time Sync Equipment clock-time alignment

Electricity I: Charge/discharge current and direction are base inputs for SOC integration, SOP power capability, and operating-stage recognition.

02 Model links From estimate to release

BMS estimates Cullen Count.
Temperature compensation

Platform data quality check Missing verification.
Time series and object verification

✓

Model correction/ integration Reference calibration
Physical binding amendments

Organisation SOC / SOH / SOE / SOP
And confidence between output

Confidence assessment Uncertainty quantification
Trust interval generation

Physical Object Cabinet / rack / module. Every state estimate maps 1:1 to a physical object and traces to input signals and model version.

BMS base estimate: Basic SOC/SOH estimates are based on current fractions, open-circuit voltage correction, temperature reimbursement and internal state of equipment.

· "State puzzle" for a full run cycle

View raw measurements, estimated medians, confidence intervals, reference events, and operating windows on one timeline.

Measure Estimated value (median) Confidence interval (95 per cent) Reference event/ calibration point Capacity test reference Verify Checkpoint

04 Estimate and reference validation

Align with the Cullen count reference (disable)

Capacity test points

Workspace Match Verification

Error distribution (all validation points)

Error Statistics (%SOC)

Indicators	Value
Average deviation Bias	-0.28
Middle Error	-0.15
RMSE	1.82
MAE	1.23
Standard deviation	1.80
95% split error	3.45

Error distributed with SOC

Validation principles: Validation points must come from an independent reference or controlled reference event; The same data cannot be calibrated as model input and independent validation findings at the same time.

05 Calibration triggers and closed loops

01 Trigger assessment Continued anomaly and drift
Compliance with calibration conditions

02 Controlled calibration Select Reference Data
Execution Parameter Amendments

03 Revalidate Authentication error
Validate stability

04 Compliance judgement Meets the release threshold
Entry approval

✓ Release new models New version effective
Version and effective date of record

Irregular
Save failed authentication record

→

Roll back
Restore previous stable version, preserve usability and security

06 Box / Cluster deviation and confidence monitoring

Rack-01

Rack-02

Rack-03. Attention.

Rack-04

Rack-05

Rack-06

Rack	SOC (%)	SOH (%)	Confidence 95%	Authentication RMSE	Status
Rack-01	54.8	92.9	96.1	1.61	&Normal
Rack-02	54.6	92.7	95.3	1.74	&Normal
Rack-03	53.9	89.6	88.1	2.91	Attention
Rack-04	55.1	92.4	95.0	1.66	&Normal
Rack-05	54.7	92.8	95.6	1.79	&Normal
Rack-06	54.9	92.5	95.2	1.68	&Normal

Rack-03: The confidence level was reduced to 88.1 per cent, RMSE was raised to 2.91, SOH was also significantly lower than the other Racks in the cabinet, and further verification of consistency between reference events and inputs was required.

07 Long-term SOH trajectory and degradation attribution

A long-term state combines cumulative throughput, temperature exposure, SOC stay, operational pressure and consistency indicators to explain degradation trends.

Degradation attribution Model extrapolation

Bulk / Loop

46%

Temperature Exposure

22%

High SOC Stop

13%

Impact of incoherence

Operating Pressure

Other

08 Governance of models and parameters

Model / Arguments Set	Version	Scope of application	Key references	Audit role	Entry into force time	Current Status	Roll back
State-Est v2.4.1	v2.4.1	LFP 280–314Ah / T ≤45°C	OCV-LUT-2025Q1 / Rint-Map-2025Q1	Algorithm Manager	2025-05-10	&Normal	But roll back
State-Est v2.3.0	v2.3.0	LFP 280–314Ah / T ≤45°C	OCV-LUT-2024Q4 / Rint-Map-2024Q4	Algorithm Manager	2025-02-18	Archive	But roll back
Degr-Physics	v2.3	LFP full life cycle	Degr-Model-2025Q1	Model Leader	2025-03-01	&Normal	But roll back

Principles of governance: Each model or parameter change must record the cause of the change, the training/calibration of the data version, the validation results, the auditing role, the time of effectiveness and the rollback target. Model upgrades do not cover the context of the version of the historical estimate.

09 Operation and quality assurance output

PDF Status report Daily/ Weekly/ Monthly

VAL Verification Records Error, deviation and reference distribution

REF Reference Event Index Still / Capacity Test List

CAL Calibration records Trigger, process and validation results

SOH SOH Track Report Trends and attribution of degradation

CI Trust and alarm. Low confidence event and disposal

Output format: PDF / CSV / Parquet / API Storage policy: Encrypted storage, version management, traceability Data integrity: 99.2%