关键词: 区间二型模糊逻辑系统, 二型非单值模糊化, BP算法, SVD-QR算法, 风电功率, 多步预测

Abstract: Aiming at short-term wind power forecasting, a method composed of interval type-2 fuzzy logic systems (FLS) with non-singleton type-2 fuzzification is proposed. Taking into account the stochastic nature of the wind power, a forecasting model using an interval type-2 FLS with non-singleton type-2 fuzzification is firstly built, the back-propagation algorithm is then used to update the parameters including the input membership function, the antecedent and consequent membership function respectively, finally, the SVD-QR algorithm is applied to the results of the BP algorithm to determine the reduced set of fuzzy rules, the training process iterates until the forecast accuracy can meet the design requirement or reach the specified training epoch. The employed method is then applied to real-world wind power forecasting instances, under the same conditions, compared to the existing forecasting methods including support vector machine(SVM), type-1 FLS, interval type-2 FLS with singleton fuzzification, interval type-2 FLS with non-singleton type-1 fuzzification, etc. Experiment results confirm that the employed method can achieve better forecasting accuracy while the fuzzy rules are reduced.

Key words: Interval type-2 fuzzy logic system, non-singleton type-2 fuzzification, BP algorithm; SVD-QR algorithm, wind power, multi-step forecasting