Present distribution infrastructure is designed mainly for uni-directional power flow with well-controlled generation. An increase in the inverter-interfaced photovoltaic (PV) systems requires a thorough re-examination of the design, operation, protection and control of distribution systems. In order to understand the impact of high penetration of PV generation, this work conducts an automated and detailed modeling of a power distribution system. The simulation results of the modeled distribution feeder have been verified with the field measurements.

Based on the feeder model, this work studies the impact of the PV systems on voltage profiles under various scenarios, including reallocation of the PV systems, reactive power support from the PV inverters, and settings of the load-tap changing transformers in coordination with the PV penetration. Design recommendations have been made based on the simulation results to improve the voltage profiles in the feeder studied.

To carry out dynamic studies related to high penetration of PV systems, this work proposes a differential algebraic equation (DAE) based dynamic modeling and analysis method. Different controllers including inverter current controllers, anti-islanding controllers and droop controllers, are designed and tested in large systems. The method extends the capability of the distribution system analysis tools, to help conduct dynamic analyses in large unbalanced distribution systems.

Another main contribution of this work is related to the investigation of the PV impacts on the feeder protection coordination. Various protection coordination types, including fuse-fuse, recloser-fuse, relay-fuse and relay-recloser have been studied. The analyses provide a better understanding of the relay and recloser settings under different configurations of the PV interconnection transformers, PV penetration levels, and fault types.

A decision tree and fuzzy logic based fault location identification process has also been proposed in this work. The process is composed of the off-line training of the decision tree, and the on-line analysis of the fault events. Fault current contribution from the PV systems, as well as the variation of the fault resistance have been taken into consideration. Two actual fault cases with the event data recorded were used to examine the effectiveness of the fault identification process.