Short term load forecasting using daily updated load models

✍ Scribed by Masatoshi Nakamura

• Publisher: Elsevier Science
• Year: 1985
• Tongue: English
• Weight: 624 KB
• Volume: 21
• Category: Article
• ISSN: 0005-1098
• DOI: 10.1016/0005-1098(85)90046-9

✦ Synopsis

This paper proposes one-day-ahead load forecasting using daily updated weekday load models and weekly updated bias models for everyday-of-the-week loads. The load characteristics are examined first for actual data from Kyushu Electric Power Company and weather stations in Kyushu throughout 1982. Then, according to properties of the loads, the algorithm of the load forecasting is derived. Features of the load forecasting are summarized as follows: (1)according to load curve properties, 24 hourly weekday load models are constructed individually during a 24-h period; (2)the weekday load models, which give estimates for the regression coefficients of weather and other factors, are updated everyday by the exponential weighted least squares method (equivalent to the steady state Kalman filter); (3)by using the bias models, the load forecasts for Saturday, Sunday and Monday patterns are also obtained by the same method used as for the weekday pattern. Based on actual data, the accuracy of the proposed load forecasting was found to be very high, the standard deviation of the relative error of the load forecast being about 3 ~.