The Relationship Between Power Load and Meteorological Factors with Refined Power Load Forecast in Shanghai

The forecast value of power load is an important reference for the power dispatch, and meteorological conditions have a significant impact on the diurnal and seasonal variation of the power load. Therefore, power load data of every 15 minutes in Shanghai and observations of Baoshan weather station (ID:58362) every 3 hours from 2004 to 2008 are analyzed to study the correlation. It's found that the meteorological power load is most closely related to the temperature. When the daily mean temperature (DMT) is great than 25 ℃, the daily mean meteorological power load rate (DMMPLR) is positive, DMMPLR increases with the increasing of DMT; while DMT is great than 18 ℃ and less than 25 ℃, DMMPLR is negative, DMMPLR increases as the DMT rises too; when DMT is between 6 ℃ and 18 ℃, DMMPLR is negative, DMMPLR decreases with the increasing of DMT; and while DMT is less than 6 ℃, DMMPLR is positive, the magnitude of changes with DMT is slightly. Moreover, the characteristics of power load diurnal variation curve display significant differences in different temperature ranges or under different weather types. Taking the summer season (T≥25 ℃) as an example, the peak of power load rate appears around 1100 BT in rainy-day, appears at about 1400 BT in the day with rainy-afternoon, and appears in the afternoon in the day with sunny-morning; the diurnal variations of hourly mean meteorological power load rate (HMMPLR) are basically the same in the day with rainy-morning or overcast-morning, suggesting that precipitation is not very important, but the sky condition has the main influence.Stepwise regression method is adopted to get the prediction equations of DMMPLR in each temperature range, and then the forecasting values of HMMPLR, under different weather types, calculated by multiplying the statistics coefficients (HMMPLR/DMMPLR) obtained in advance. The forecast test results in 2009 show that, using the 3-day average (before the forecast date) of the trend power load as the trend power load of the forecast date, the mean of absoulute relative error (MARE) of daily mean power load forecast value (DMPLFV) is about 3.6%. The MARE of DMPLFV of non-working days is larger than that of working days. In working days, while DMT is greater than 18 ℃, the MARE of DMPLFV is lower, when DMT is less than 18 ℃, the MARE of DMPLFV significantly increases. The MARE of hourly power load forecast value is about 4%.