Short-term load forecasting for energy markets

The problem of the application of neural predictors for Short-Term Load Forecasting (STLF) for energy transactions planning in utility is discussed. The conditional expected value of the energy demand approximated by the neural network forecast does not necessarily indicate the optimal size of the order, according to the financial regulations associated with energy markets. Two approaches to the solution of this problem are presented. The first one is based on GANN - hybrid neural-genetic STLF system, trained with the financial cost function. The second approach rely on determining of the sources of uncertainty for classical neural predictor.