ChinaEnergy and utilities2025 – ongoing

Guangxi Power Grid

AI transmission inspection and dispatch support

Guangxi Power Grid reported analysing 20 million inspection images with AI

WorkingCompany's own claim

Problem

Mountainous terrain made transmission inspections difficult and exposed workers to hazards. Variable renewable generation also complicated dispatch.

Approach

Guangxi Power Grid combined autonomous drone inspections with a computer-vision model built on China Southern Power Grid's Dawatt foundation. It fed missed and incorrect detections back into model training. Separate forecasting and dispatch models supported grid operations.

Results

20 millioninspection images analysed cumulatively, according to the utilityCompany's own claim[1]

up to 46transmission defect and hazard types the model could identifyCompany's own claim[1]

over 80%reported defect and hazard identification accuracyCompany's own claim[1]

Takeaway

Inspection automation still included human review and retraining from detection errors. The reported accuracy did not establish how many defects the model missed.

Sources

人工智能开启广西电网数智变革新图景
China Energy News22 Sep 2025News article

Source summaries

[1]China Energy News22 Sep 2025

The report described operational inspection, forecasting and dispatch systems. It relayed utility figures of 20 million images analysed and identification accuracy above 80%.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026