| Keywords |
fault detection; Optimal hyperparameter; CUSUM model; PV generation system |
| Abstract |
This study propose fault detection algorithm of photovoltaic(PV) generation system for efficient operation and management(O&M) with cumulative sum control chart (CUSUM). The fault characteristics was analyzed to operation data between inverter and PV power output in representative day of fault. Each current of channel compared with PV power output to identify minor faults that could not be found in the inverter. Finally, fault detection algorithm tested to find optimal hyperparameter with drift and threshold in channel of PV generation system. As a result of the characteristics between operation data (current, voltage, input power, power factor, module temperature, solar irradiation) and PV power output, fault type of current, voltage, input power, and power factor are identical to PV power output at fault sections. Also, whole channel showed fault characteristics at same sections. However, some channels showed normal pattern in other fault sections, so it is limited to identify fault type at inverter and connection band. Optimal hyperparameter of the CUSUM was searched to detect fault with different fault level of the channels. In case of threshold=0.3 and drift=0.005 of CUSUM, the model demonstrated high detection performance in decreased by 50% of current. The threshold=0.25 and drift=0.01 was optimal hyperparameter in decreased by 20% of current. The results show that CUSUM not only accurately detects fault section but responded some weather changes as normal with optimal model condition. |