| I&T Solution |
AI tool for Automatic Fault Detection and Diagnosis (AFDD) of photovoltaic (PV) systems
(REF: S-1605) |
| Trial Project |
|
| Solution Feature |
- Python-based self-adaptive AFDD model for PV site energy output prediction that is web-embedded and is designed to accept the large amount of existing EMSD and other 3rd parties solar plant’s historical data for AI training
- The resulting AFDD model will greatly ease the needs of typically years of on-site data feeding and allow immediate satisfactory energy output prediction accuracy for a given PV site with only limited target data
- Detect likely-failing PV site equipment by using the AFDD’s automatic compound-fault detection and diagnosis capabilities
- Able to combine available PV panel DC outputs and inverter fault code data to further enrich the details of AFDD’s fault prediction, e.g. able to tell which PV string is ageing earlier than others
- Ability to continuously increase the AFDD’s self-adaptiveness and generalization level, by allowing AI training with self and 3rd party PV site data. This greatly eases specific AI-model fine-tuning and huge development costs when new sites are added
|
| Trial Application and Expected Outcome |
- Web user will get on-line real-time PV equipment condition assessment report that is based on systematic judgement between actual and AI-predicted PV site performance via the establishment of: - a real-time detection system for key electric parameters - an intelligent decision-making support system with knowledge graph
- PV site operators will get PV equipment pre-matured fault alarms to indicate system abnormality at early stage to minimize the system downtime with less manpower
- Based on the AFDD model reported conditions of PV equipment, the system manager can plan condition based PV predictive maintenance strategy
- Possible to effectively transfer existing learning-based deep-neural-networks AFDD model for fault classification to PV sites that have insufficient historical data, hence enable immediate PV equipment assessment for new sites
- Ability to extends AFDD capability by adding on-site video camera with automatic image-based fault detection algorithm to detect and verify fault e.g. physical damage of PV panels and damages to the supporting structure
|
| Additional Solution Information |
PV Investigator - AFDD - Trial Application and Expected Outcome - v2.pptx
|
| Info on I&T Solution Provider |
| Solution Provider | : | GDEPRI Power Control Systems & Equipment (HK) Limited | | Address | : | Room 1909, 19/F., Tai Yau Building, 181 Johnston Road, Wanchai, Hong Kong. | | Contact Person | : | Stanley K W Leung |
| Position | : | General Manager | | Tel | : | 94628258 | | Email | : |
stanley.kw.leung@gdepri.com | | Webpage | : | www.gdepri.com |
|
