On-line short-term bus load forecasting for distribution substations

Short-term bus load forecasting functions in the distribution substations are determined. The appropriate in the choice of the stochastic models connected with specific features of the load are presented. For load modelling Box Jenkins time series analysis and ARIMA models are used. The autocorrelation and partial autocorrelation functions of the raw load data values and differencing one are used for preliminary identification of the model. Based on the above, the appropriate ARIMA model is chosen. A comparative study with an alternative model according to data of the load forecasting is presented and it is shown that the pure AR model is more appropriate. In order to improve its features, the model is extended, using additional intervention variables such as trigonometric polynomial. The additional means to improve estimation and forecasts by the adaptive technique are described. Identification of individual ARIMA models for workdays and holidays is recommended for future development.