Forecasting The All-Day Curve Of Summer Electric Load With Weather Sensitivity

Forecasting the electric load is an important research field with practical sig-nificance. Aiming at many problems of the grid aroused by the summer electricload, such as cutting or restricting the electricity due to being short of the peakload, and overloading which makes the grid to run insecurely and unsteadily, inthis paper we research some problems in the field of short-term load forecasting, in order to forecast the all-day curve of summer electric load with weather sen-sitivity. On the basis of summarizing some traditional and uptodate methodsin the field of forecasting electric load, a mixed method to forecast the all-dayload curve is presented, which integrates the data mining and analysis of timeseries. It beforehand finds and predicts the stable point of load curve, accordingto their distribution on the 24 segments of the load curve, and treats these seg-ments by different means, including some data mining techniques of time series, or the ARMA(p, q) model of time series analysis. In the light of the fact thatload value of a point can't be much different from its neighbors, the method alsosmoothes the load curve according to the rules which are shifting the stable pointas little as possible and least adjusting the whole curve, which can be modeledas a linear programming problem. Besides, using the load and weather data ofsome city, we have tested this method of forecasting all-day load curve.