Interval prediction of ultra-short-term photovoltaic power based on a hybrid model

Abstract

Accurate photovoltaic (PV) power prediction facilitates large-scale PV grid connection and ensures the safe and reliable operation of power system. However, the information obtained from traditional point forecasting results is limited and not reliable enough for decision-makers to execute operational strategies. To obtain more abundant forecasting information than the point prediction method, we develop a hybrid interval forecasting model by combining fuzzy information granulation (FIG), improved long and short-term memory network (ILSTM) and differential autoregressive moving average (ARIMA) model to predict the interval of PV output power. First, the original PV power is decomposed into some components by complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), then the new sequences are obtained by fuzzy entropy (FE). Second, FIG is applied to mine the reconstructed sequences and obtain the maximum, minimum and average value. Third, the high frequency components are predicted by ILSTM, while low frequency components and residual are predicted by ARIMA. The final interval prediction result is obtained by summing up the interval prediction results of each component. The performance of the proposed model is verified by comparing with other models. Experimental results demonstrate that the presented model can accurately cover the actual PV power values and deliver more valuable decision information for power system dispatch.