Show simple item record

dc.contributor.authorAruho, Karugahi Joseph
dc.date.accessioned2017-09-13T00:49:52Z
dc.date.available2017-09-13T00:49:52Z
dc.date.issued2013-09
dc.identifier.citationAruho, K. J. (2013). Short-term load forecasting for Uganda’s transmission network. Unpublished master's thesis, Makerere University, Kampala, Uganda.en_US
dc.identifier.urihttp://hdl.handle.net/10570/5707
dc.descriptionA dissertation submitted to the Directorate of Research and Gaduate Training in partial fulfillment for the Award of Master of Science in Electrical Engineering Degree of Makerere Universityen_US
dc.description.abstractThis research attempts to simulate and develop by use ANN, test, and recommend a reliable and accurate model of forecasting the next day’s electrical load demand for UETCL in Uganda. UETCL is the sole purchaser from UEGCL and most of other producers (generators), and dispatcher to (UEDCL- Umeme) and some other neighboring countries of the electrical energy/load in Uganda. Short-term electrical power load forecasting at UETCL is solely done to meet the demand and is not accurately done. Only the maximum value of historical load demand from the four similar previous consecutive days is chosen as the forecast. Yet modern, short-term forecasting methods consider other factors like weather (temperature and humidity), time factors (minute of the hour, time/hour of the day, day of the week), electricity prices, seasonal (Christmas, international football) and customers’ classes (residential, commercial & industrial) to forecast the load demand. There are a number of benefits and significance of accurate load forecasting as highlighted in the text. The research presents the development of an Artificial Neural Network (ANN)-based shortterm load forecasting (STLF) model with improved accuracy. By use of regression analysis in the STATA and ALYUDA FORECASTER (ANN) packages of software, factors that significantly affect the demand were determined and applied, the model developed by simulation of ANN and its accuracy compared with the actual and forecasted demand at UETCL. The proposed ANN is trained and tested with weather-related data, special events, time related factors and historical electric load-related data using the data from the period November 2011, to April 2012. The model’s forecasting is based on the fact that the ANN develops its own model by learning examples through training given domain data (significant factors and actual demand) and the global solution depends on the least average MSE, highest R2 and correlation coefficient. As far as the development of the model is concerned, 83.08% of the actual demand data is explained by the significant determining factors data and 91.41% of the independent variables data is correlated to the actual demand data. This entails a good fit of the model. The least average MSE obtained was 319.92. By considering the daily’s total demand of one month of April 2012, the model’s percentage error ranges from -12.32% to +3.57% whereas that of UETCL ranges from -20.66% to +0.47%. For the month of April 2012; the minimum, maximum and average MAPE values obtained for the model are 3.43% Day14 (SAT), 18.85% -Day4 (WED) and 8.38% while those of UETCL are 4.05% -Day28 (SAT), 24.79% -Day4 (WED) and 11.26% respectively. The model’s RMSE values obtained re 15.16, 66.16 and 30.73 whereas those of UETCL are15.52, 83.2 and 36.51 respectively. These error figures are comparable with best models and practices around the globe. Load forecasting has become, in recent years, one of the major areas of research in electrical engineering. Most traditional forecasting models and artificial intelligence neural network techniques have been tried out in this task. ANNs have lately received much attention, and a great number of papers have reported successful experiments and practical tests. Despite spending millions of money, the simple model developed by simulation of ANN in this research, could provide utility decision-makers with cost effective and timely answers to many strategic questions concerning electricity supply and transmission management. The next logical step for the research is to improve further the performance of ANN, for this model, perhaps through better training methods, or better input.en_US
dc.language.isoenen_US
dc.publisherMakerere Universityen_US
dc.subjectTransmission networken_US
dc.subjectElectrical load demanden_US
dc.subjectPower load forecastingen_US
dc.subjectElectricity pricesen_US
dc.titleShort-term load forecasting for Uganda’s transmission networken_US
dc.typeThesis/Dissertation (Masters)en_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record