A new short-term load forecasting approach using self-organizingfuzzy ARMAX models

摘要

This paper proposes a new self-organizing model of fuzzynautoregressive moving average with exogenous input variables (FARMAX)nfor one day ahead hourly load forecasting of power systems. To achieventhe purpose of self-organizing the FARMAX model, identification of thenfuzzy model is formulated as a combinatorial optimization problem. Thenna combined use of heuristics and evolutionary programming (EP) scheme isnrelied on to solve the problem of determining optimal number of inputnvariables, best partition of fuzzy spaces and associated fuzzynmembership functions. The proposed approach is verified by using diversentypes of practical load and weather data for Taiwan Power (Taipower)nsystems. Comparisons are made of forecasting errors with the existingnARMAX model implemented by the commercial SAS package and an artificialnneural networks (ANNs) method