自動標本選択による週間電力負荷予測

DOI Web Site Web Site

書誌事項

タイトル別名
  • A Method for One-day-through Seven-day-ahead Electrical Load Forecasting with Automatic Sample Selection
  • ジドウ ヒョウホン センタク ニ ヨル シュウカン デンリョク フカ ヨソク

この論文をさがす

抄録

An approach to one-day-through seven-day-ahead electrical load forecasting is proposed for the purpose of weekly power station operation. The load is affected by various factors. However, it is impossible to consider all the factors in load forecasting; the relationship between the load and some factors are not clear; and sometimes their own forecasts are required. In our case, information about day of the week and the temperatures obtainable from weekly weather forecasts can be the only reliable input variables to the forecasting model. The other factors which are not included in the model occasionally make the load significantly different from the model output. The data samples measured in such situation reduce the reliability of the model and thus the forecasts. In order to obtain reliable forecasts we need to eliminate these samples out of the estimation procedure of the model, and investigate the reason of such large errors to improve the forecasts. In this paper, an automatic sample selection method is proposed: each measured sample is judged, based on the estimation error, whether it is appropriate for the model estimation or not. With the method we can effectively eliminate harmful samples without loosing useful information contained in the data. The procedure for finding out which factors cause the large errors and how they do is also discussed as well as the method for correcting the forecasts. Some forecasting examples show that we can forecast the load with accuracy of 4 percent average error. Some useful information is also derived from the results which explains large errors and will serve to elaborate the forecasts.

収録刊行物

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ