What the data for this category looks like
Data sources for energy storage yield primarily include local power station monitoring systems, regional power trading platforms, and grid dispatching interfaces. Data updates occur once daily. Full data for the previous day is collected by early morning the next day.
Each data entry includes fields such as unique power station identifier, total installed capacity, total daily charge and discharge power, peak-valley time period electricity price range, operation and maintenance cost, daily net profit, and cumulative annual profit. All units follow uniform standards: capacity is measured in kilowatt-hours, electricity price is measured in yuan per kilowatt-hour, profit is measured in yuan, and no additional composite units are used.
Constraints for Model Access and Configuration
The daily update rhythm of energy storage yield data requires that model access be configured with a fixed daily scheduled trigger task. Requests must adapt to the T+1 data delay characteristic.
The presence of peak-valley electricity price range fields requires that multi-interval parameter parsing rules be configured during model access. This supports extraction and calculation of continuous value ranges.
Minor field differences exist across different energy storage power stations. The dynamic field adaptation mode of the model must be enabled to avoid call failures caused by missing fields.
Cross-platform data source call requirements require that unified interface authentication rules be configured. This adapts to signature verification logic of different data sources.
In scenarios with large daily data volume, the model's batch request threshold must be adjusted to avoid timeout during single calls.
How to Set Configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
modelTriggerCron | 0 0 2 * * ? | Adapts to the T+1 update rhythm of energy storage data, triggers calls at 2 AM the next day to ensure retrieval of the latest previous day's data |
dynamicFieldEnabled | true | Minor field differences exist across different energy storage power stations. Enabling dynamic adaptation avoids call failures caused by missing fields |
batchRequestThreshold | 50 items per request | Daily energy storage data volume is large. Adjusting the batch request threshold avoids timeout during single calls |
maxRequestTimeout | 600 seconds | Cross-platform data pulling and parsing take a long time. Extending the timeout period avoids call interruptions |
apiAuthType | signature | Adapts to cross-platform authentication requirements of energy storage data sources, unifies interface verification rules |
embeddingModel | text-embedding-3-large | Matches common index model requirements, supports vector generation for long text |
The parameter values provided on this page are common starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis. It is recommended to test on your own samples before finalizing settings.
Three Common Mistakes
- Scenario: A separate model key configured in OneAPI triggers a null value error when entered into FastGPT, with the interface returning a 401 status code. Cause: Only the key was entered, and the basic interface address of OneAPI was not configured, leading to an incorrect request path that failed authentication verification.
- Scenario: In FastGPT V4.9.6, the OneAPI-specific settings option does not appear in the model configuration interface. Cause: This version does not include a built-in OneAPI quick access template. The basic URL and authentication parameters of a custom model must be entered manually.
- Scenario: After configuring
text-embedding-3-largeas the index model, the data indexing task gets stuck with no response. Cause: ThebatchRequestThresholdparameter was not adjusted. The energy storage data volume of a single request exceeds the maximum input length supported by the model, leading to task blocking.
How to Confirm Successful Configuration
- Manually trigger a model call, and check whether the returned results include all fields of energy storage yield, with no missing or parsing exceptions.
- Check the scheduled task execution log to confirm whether the scheduled daily call task starts normally and completes full data pulling.
- Verify the vector generation process of the index model, confirm that corresponding vectors can be generated normally after entering energy storage data fields, with no blocking or errors.
- Cross-check the interface authentication log to confirm that the FastGPT request parameters fully match the authentication information configured in OneAPI.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.