Most of the electrical power of Ethiopian cities has fed by a radial distribution system. Mekelle is one of the cities supplied by radial network. The case study in this thesis is applied in Mekelle power distribution lines specifically in feeder line KO5.The feeder line is modelled as a single network that has 112 nodes and 64 loads. It has a total capacity of 10.5 MW. The voltage profiles of most of the buses in feeder line are not in the acceptable voltage level (i.e. 0.95-1.05 pu interval). Moreover, the active and reactive power losses of the feeder are 1498.7637kW and 1481.461 kVAr respectively. Therefore, voltage profile improvement and power loss reduction improvement have needed to this problem. Distributed generations are selected as the best solution. A novel method-loss sensitivity based Newton Raphson-Genetic algorithm approach have implemented for sizing and placement of the DG. This method places one DG at a time on a high loss sensitive bus. After that, it calculates the size of the DG corresponding to that bus. For the next DGs, same practice continues until the specified stopping criterion has reached. In this thesis, for comparison purpose load flow analysis for different scenarios has done. It has been studied the impact that distributed generation has on the system. The scenarios have considered based on the type of DGs. PQ and PV type DGs have considered. The ratings of the PQ type and PV types DGs are tested according to the genetic algorithm best fitness function. Thus, single PQ type DG has selected as the best solution to optimize the system. As a result the total active power loss recorded in feeder has minimized to 145.1 kW and the total reactive power loss has minimized to142 kVAr . In the same way all the weak voltage profiles have also optimized to the standard ±5% voltage deviation level.