PV O&M Optimization by AI Practice |運用人工智慧實現案場維運的最佳化


Mr. Maoyi Chang  / 張茂益

Vice General Manager ( O&M BU)






  • Vice General Manager, O&M BU, Sinogreenergy
  • Manager, O&M Dept., Sinogreenergy
  • Manager, PV Quality Engineering Dept., AUO
  • Manager, PV Reliability & FMA Lab., AUO (UL WTDP Lab. for IEC 61215, 61646, & 61730)
  • Speaker & Oral Author, EU PVSEC (2014, 2015, 2019)   
  • Oral Author, IEEE PVSC (2015)
  • Owned 14 US patents & 15 TW patents



  • Master of Material Science and Engineering, National Tsing Hua University




To deploy solar power as the reliable renewable energy source in massive scale, implanting efficient operation and maintenance (O&M) acts as a crucial role play. we demonstrate utilizing artificial intelligence (AI) technology to optimize O&M task for more than 150 project sites up to 54MW with 11 inverter brands and 9 module suppliers simultaneously. Without specific module, inverter and location parameters as inputs, the power prediction model for each inverter-MPPT is trained based on its own historical production data and established as its own fingerprint. Each inverter-MPPT behavior from all the projects (over 7,583 MPPTs) is monitored and analyzed by machine learning in every five minutes. Fault detection alert with failure mode is automatically judged, and prompt notification is sent to user by mobile device or email. From one-year implementing result, the average energy yield of 0.16 kWh/kWp/day (4.7%) is improved for the project sites with this AI system than those without. The precision for fault detection is 99.2%, and it also reaches 92.3% for failure mode diagnosis. Taking the advantage of the high accuracy of fault detection and failure reason diagnosis, the efficient and prospective O&M would be achieved.



Share page with AddThis