Load forecasting plays an exceptionally important role when solving a wide range of problems associated with control planning and design of power systems and subsystems. This paper describes approach based on combined using methodology of fuzzy sets and neural networks to solving problem of load forecasting. Application of unsupervised/supervised learning concept is realized in proposed algorithm. A new procedure to clustering patterns of training set which does not need iterate computation is proposed. Two algorithms oriented on using fuzzy (including linguistic) as well as interval initial data are considered. Proposed approach is well concerted with real level of uncertainty of initial information in power systems and subsystems and provides adequate and quite accurate solution of load forecasting problem.
展开▼