An efficient approach for short-term substation load forecasting

Load forecasting methods for a large geographical area such as New England are widely established. However, substation load forecasting is much more difficult, since load patterns of substations are more irregular than that of a large system. In addition, considering the large number of substations in an area, computation time is also an important issue when forecast all the substations together. In this paper, an efficient approach is presented for short-term load forecasting of all substations within a given system. The key idea is the addressed load pattern similarities analysis between substations and zone which is upper grid than substation. If load patterns of substations are similar to that of zonal load, forecasting of these substations can be directly obtained from proportion of the zonal forecasting results. For those substations whose load patterns are different from that of the zonal load, artificial neural network is used to capture the complicated substation load features. Numerical testing of the presented method demonstrates the effectiveness of our method based on 23 substations within two zones.