Daily load forecasting based on previous day load

Abstract

In this paper we consider daily load forecast problem and explore the idea that similar conditions to those at the forecasting moment have normally existed. before. If the load conditions change relatively slowly, then the yesterday's load curve can be used as an indicator of the load conditions of the present day; so it is assumed the robustness of the model. To test the idea of the robustness two models are considered. The first model uses the self-organizing map (SOM) to form network weights. The map is trained on the load data of ten months. The forecast is received by connecting load data of the previous day to a weight vector that contains a forecast for the target day. The second model that we suggest here is a considerable simplification of the first one and is based on the idea of the nearest neighbor.