A novel approach for load profile formation and technical loss estimation of electrical disrtibution systems

The importance of having knowledge on the characteristics of an electrical distribution system demonstrates itself as a key factor to how a system shall be operated and how new investments shall be made. The characteristics of a distribution network depends heavily on parameters such as the types of different equipments that are utilized, whether that specific network is radial or closed and classification and consumption rates of various kinds of end-users. In addition to these parameters, load profiles of endusers which are subscribed to the network are also a crucial determinant factor. Generated electrical energy is directed to end-users via transmission and distribution networks. It is an apparent fact that there is difference between generated and consumed electrical energy. This difference yields itself from two types of losses named technical and non-technical losses. Technical losses result from the losses occur in the equipments which are part of the distribution network. This type of loss which depends on the flowing current and the voltage, varies based on the load profiles of end-users. Composition of non-technical losses consists of non-chargeable or non-measurable electrical energy. In the process of calculating current-based technical loss which is also known as I2*R loss, knowledge of load profiles retains a key role. It is significant to reckon the fact that two different end-users which have the same kWh consumption rate might have totally different load profiles which in the end create two seperate and highly differentiated technical loss characteristics. There exists two different methods used for calculating technical losses. The first method is based on load and load loss factor dependant calculation which comprises calculations derived from total and maximum power measurements. Although being a very practical method, this load and load loss factor based calculation harbors increased probability of being erroneous. The second method used for calculating technical losses consists of calculation utilizing load profiles. It is based on the calculation of the area under the load profile curve which is formed by minute and hourly consumption data. This second method which is based on load profile curves yields more accurate and closer result to actual loss rate. Different types of end-users obviously have different 24-hours of load profiles. The exact time or time interval on which a very specific end-user reaches its maximum load rate may vary according to seasonal, daily or hourly needs of utilization. These mentioned factors which result from reaching to maximum load rate at different times directly create the notion that different types of end-users affect the distribution network at very distinct and discrete patterns based on their specific load curves. To illustrate, a street-lighting pole reaches its maximum load in night time whereas a work office reaches its maximum load during day time. This distinction creates many different load profiles with different maximum load time or time interval, stating that it might be very hard to estimate different load profiles. The best way to estimate the characteristics of a distribution network is to allocate smart meters to end-users in order to obtain load profiles. However this can't be achieved easily due to high investment costs. Therefore the distribution network has to generate a smart way to estimate load profiles. Four different clustering methods are widely utilized whilst forming load profiles. These methods are K-mean algorithm, Fuzzy C-mean algorithm, hierarchical clustering and PCA method. Mentioned methods are often used when there is a lot of unknown data. Moreover, due to grounding their bases on a very error-prone and complicated reasoning which is purely consumption rate of an end-user, these methods provide insufficient and ineffective means of estimating load profiles. Expanding on this claim, 24-hours operating factory though having the same monthly consumption as a shopping mall, has very different load profile curve. The factory consume almost constant electrical energy throughout the day and night time whereas shopping malls cannot consume that much in the night time due to its closure. In addition to this, above mentioned four methods are not feasible enough to rely on. This work provides new solution by categorizing end-users according to their consumption characteristics and creates clusters having their own unit-profile. Two different end-users in the same cluster which possess different energy consumption rates, have similar load profile curves whereas other two end-users belonging to different clusters might have the same energy consumption rates but there exists a very low probability that those two end-users possess the same load profile curves. Four methods mentioned above paragraph rely not on load profile curves but on total consumption rates, creating a very erronous approach. Arising from these reasons, this work demonstrates a new approach on how to cluster end-users, based on their load profiles. Furthermore, these formed clusters can be divided into sub clusters with the proposed new approach supplied by this work. The base construction material of this work is creating a new approach based on real set of consumption data gathered from a distribution network. Four different clusters named industry, household, commerce house and lighting have been assigned according to their specific load profile curve characheristics. Thereafter commerce house cluster is further divided into nine sub clusters such as general store, bank and so on. Industry cluster is then also expanded into three different sub clusters according to their operating hours and production shifts. These main clusters and divided sub clusters are examined in detail by their consumption on yearly, seasonal, monthly, daily and hourly basis. With the help of detailed load profile curves, each cluster and sub cluster give away their own fundemental characteristics. This work not only set its sight on bunch of data set which explains occasional consumption characteristics on time basis but also penetrate deeply into comparisons between members of clusters and sub clusters, fundamentally based on working or operating hours of each end-user in order to verify the need why such distinction between various clusters and sub clusters have to be made. The detailed analysis of similarities or differences that various end-users belonging to different clusters or sub clusters is executed. Unit load profiles have been assigned for clusters and these unit profiles are scaled according to consumption rates for the sake of creating a calculator tool software which calculates load profile of an end-user, a cutout box, a feeder or a distribution transformer. With the help of calculator tool software developed in this work, load profile estimation of a real life distribution transformer zone, which accomodates cutout boxes of above mentioned end-users from different clusters and sub clusters, for two different cases is executed and then compared to real load profile data obtained from allocated smart meters. First, by the use of consumption rates of end-users from different clusters in that specific distribution transformer zone, load profiles have been calculated. Calculated load profiles belonging to different clusters are then utilized to estimate load profile and technical losses on a distribution transformer zone in December 2015. This analysis is done with the aid of a power system analysis software which is capable of executing load flow analysis. Second, by the use of consumption rates of end-users from different sub clusters in that specific distribution transformer zone, load profiles have been calculated. For commerce house cluster, the new estimation would require not only just the cluster itself but also sub clusters with their own data. For the industry cluster the new estimation would also require the sub clusters with their own data. Calculated load profiles belonging to different sub clusters are then utilized to estimate load profile and technical losses on a distribution transformer zone in December 2015. This analysis is done with the aid of a power system analysis software which is capable of executing load flow analysis. Finally, estimations made by using proposed new approach are compared with the real load profile data gathered from smart meters in a distribution tranformer zone for December 2015 and the final result is examined in detail. Estimated technical losses for the same consumption rates by just considering main clusters when compared to the real data, exhibits just 2.23% deviation. Estimated technical losses for the same consumption rates by considering sub clusters when compared to the real data, exhibits just %1.9 deviation. Calculations and load flow analysis conclude that by just considering main clusters, the new approach yields a very accurate estimation. However expanding clusters into sub clusters and adding each load profile of sub clusters into estimation provides even better results. The difference of 0.33% between cluster based and sub cluster based estimation proves expansion to sub cluster brings the new approach closest to the real data. In any case, both estimations are definitely so close to the actual values. The extent of this work provides assistance to observe how important it is to divide main end-user clusters into sub clusters whilst forming load profile curves in order to provide closer estimation to real consumption data. This work also provides numerical data to verify each observation and estimation. Moreover, this work also provides a completely feasible, more accurate when compared to other clustering methods and more utilizable calculator tool which might be used by a distribution network in order to estimate load profiles of an end-user, a cutout box, a transformer and a feeder.