02780nam a2200241   450000500170000000800410001703500150005810000190007324500620009226000520015430000350020649000270024150001050026850200580037352017670043165000480219870000360224670000410228270000490232371000420237281000590241485600650247320260818114246.0080998         th                  eng    a.b114930331 aLee, Yean-shin10aForecasting of power distribution load in Tainan district  aBangkok :bAsian Institute of Technology,c1995  a88 leaves +e1 online resource1 aThesis ;vno. CS-95-12  aA thesis submitted in partial fulfillment of the requirement for the degree of Master of Engineering  aThesis (M.Eng.) - Asian Institute of Technology, 1995  aAlthough considerable work has been done in the past on the application of some type of systematic approach to electric power forecasting, its application to distribution load forecasting has unfortunately been somewhat neglected, especially the short-term distribution load forecasting for only certain areas like Taiwan Power Company (TPC) Tainan District. Recently years, due to the uprising of environmental protection consciousness, the land acquisition of not only power plants but also secondary substations becomes more difficult than before. Most Districts of TPC will face the embarrassment of power shortage and area rationing of power commodity predictably will be seen more often. In these circumstances, reliable short-term distribution load forecasting tools will enable TPC to plan ahead of time for peak demand and inform their customers as earlier as possible when area rationing is inevitable. Three forecasting approaches namely Neural Network, Box-Jenkins and Winters were investigated through the study cin distribution daily peak load and weekly peak load forecasting for TPC in Tainan District. The results showed that the daily peak load forecasting performance by Box-Jenkins model along with standardization method is the best among the approaches considered. For weekly peak load data the Winters' method gives the best forecasting performance with lead times from one to three weeks. For daily peak load data, the accuracy of forecasting with Box-Jenkins seasonal models is much better than that of non-seasonal models. Moreover, with standardized daily peak load data the forecasting performance is improved significantly. The Backpropagation Networks can not give a good forecasting performance when only time series data are used.10aElectric power-plantsxTaiwanxTainanxLoad1 aHuynh, Ngoc Phien,eChairperson1 aDo, Ba Khang,eExamination Committee1 aNagarur, Nagendra N.,eExamination Committee2 aR.O.C. Government,eScholarship donor2 aAsian Institute of Technology.tThesis ;vno. CS-95-12  3Full-Textuhttp://203.159.5.9/ait-thesis/detail.php?q=B14996