Modeling the Adaptive Uncertainty Sets of Robust Optimization Based on Long Short-Term Memory Network and Modified Fuzzy Information Granulation

• Main Authors: Yibin Qiu, Qi Li, Yuru Pan, Lanjia Huang, Cai Sun, Hanqing Yang, Jiawei Liu, Weirong Chen
• Format: Article
• Language: English
• Published: IEEE 2020-01-01
• Series: IEEE Access
• Subjects: Robust optimization; adaptive uncertainty sets; long short-term memory network; modified fuzzy information granulation
• Online Access: https://ieeexplore.ieee.org/document/8972375/

To better balance the reliability and conservativeness of uncertainty sets of robust optimization, the concept of adaptive uncertainty sets is proposed in this paper. There are two processes contained in the proposed adaptive uncertainty sets, which are point prediction and uncertainty sets determination. In the process of point prediction, the Long Short-term Memory Network (LSTM) is used to predict the renewable energy output. In the process of uncertainty sets determination, firstly, the prediction data is granulated based on the Modified Fuzzy Information Granulation (MFIG). Then the adjustable parameters are introduced to modify the upper and lower limit parameters of the information granules. Based on the above, the modeling of adaptive uncertainty sets can be achieved. To verify the performance of the proposed adaptive uncertainty sets, three groups of wind power output data of California are introduced to the contrast experiments. The simulation results demonstrate that, under 90% confidence level, the adaptive uncertainty sets method has a higher prediction interval coverage probability and a smaller prediction interval average width compared to the box uncertainty sets and the ellipsoidal uncertainty sets, which illustrates the good performance of the adaptive uncertainty sets in reliability and conservativeness.