Distribution reliability improvement through optimal location of load break switches in 33 kV network

Abstract

Distribution network is responsible for many of the interruptions in an electricity supply system. This is internationally the case due to the more exposed nature of the distribution lines which are mostly of overhead type going through different zones like vegetated areas, coastal areas, urban areas, flat fields and hilly areas. The causes for failure as well as the failure frequency differ from one zone to the other. In Sri Lanka there are 33 kV lines of 25,257 km length spread all over the country going through different landscapes. The lines are all radial and a failure at the beginning of the line lead to supply interruption for all the connected customers. However, a failure towards the end of the line if properly managed would interrupt only a section of the customers. This paper presents a method to decide on the best number of load break switches and the optimal location of each of the switches considering nature of the different zones the line is going through and the distribution of the customers. Feeder failure data has been gathered and used to estimate failure rates corresponding to different exposure zones using the least square fit. Failure rates so estimated are used to model the failure behaviour in a given feeder going through a combination of exposure zones. Each feeder is divided into a series of logical sections based on the location of distribution transformers serving customers. If there are n such sections r (?n) breakers can be located in nCr different ways and for each of such constellations a cost function is evaluated and the System Average Interruption Duration Index (SAIDI) is calculated using number of customers that are isolated by each of the switch and the average restoration time. The total unserved energy can also be calculated. The cost function gives the sum of cost of unserved energy due to interruptions and the cost of breakers on an annual basis. Optimum arrangement can be obtained for given r and known n. By repeating this optimization for all possible r, the global optimum is obtained. This algorithm has been effectively implemented.