Study On Transformer Load Forecasting Method

Currently, in the field of power system load forecasting, the researcheson load forecasting methods are basically for the total system load demand,but lack of researches on the transformer load forecasting method. However,in the actual operation of the power grid, merely getting the prediction resultof the total regional load is not enough. While arranging the operation mode,load cutover plans, load-peak averting plans, and the new power stationcommissioning plans, the electric power dispatching department needs toknow the specific load status of each main transformer, so as to predict thespecific location or the other information of the load hot spots.Against the particularity of the main transformer load forecasting, theload forecasting methods for main transformer are studied in this paper. Themain research contents and the innovation points of this thesis include:The major research contents and results of this thesis are as follows.(1) Take the Shanghai Songjiang distribution network as example, thechanging characteristics of the main transformer load trend were analyzed.The important influential factors of the main transformer load trends werestudied, and the particularities of the main transformer load forecast againstthe regional load forecast were summarized.(2) The application of the seasonal time series methods on the main transformer load forecasting was discussed, and then proposed a methodbased on decision tree which could automatically select the best algorithmdepending on the historical load trend of transformers. The experiment resultshows that the method worked well in classification.(3) For the seasonal load cutover issues between the main transformers,a load transfer coupling (LTC) method was introduced, and an extended ap-plication of this method was done to solve the load cutover event amongmultiple transformers for more than once. Tests had been done to verify theeffect of the method.(4) For the non-seasonal load cutover issues between the main trans-formers, a transformer load forecasting method using least squares fitting wasproposed. A series of tests have been done to the110(35) kV main trans-formers in three functional areas of Shanghai Songjiang District. The resultshows that the method is superior to the seasonal time series forecastingmethod in the prediction effect, and is not necessarily informed of the exactload switching history of the main transformers. Only three years of historicalload data are required to obtain a fine prediction result, which is practical toresolve the load forecasting problem of the main transformers involved in theload switching events.