Prediction of energy consumption and risk of excess demand in a distribution system

摘要: Abstract An empirical model for prediction of energy consumption in a distribution system is described. The resembles normalized radial basis function neural network whose neurons contain prototype joint data about the process and environment. A set patterns environmental variables formed from record multi-component time series by self-organized process. Prediction performed conditional average estimator based upon known given future values variables. Importance these determined genetic algorithm. performance tested on one-year-long gas system. error difference between predicted actually observed consumption. Its value depends amounts to few percent actual probability estimated properly selected interval prediction. This can be used estimate risk demand beyond some prescribed value. For an optimization process, cost that includes operation control costs as well penalties related excess proposed. minimum corresponds economically optimal distribution.