Velocity Predictions Research and Prediction Horizon Determination. (May 2017)

• Record Type: Journal Article
• Authors: Kong, Zehui; Zou, Yuan; Liu, Teng
• Is Part Of: Energy procedia. Volume 105(2017)
• Page Start: 3616
• Page End: 3622
• Publication Date: 2017-05
• Subjects: Energy management -- Velocity prediction -- Reinforcement learning -- Hybrid electric tracked vehicle; Power resources -- Congresses; Power resources -- Periodicals; Power resources; Conference proceedings; Periodicals; 333.7905
• DOI: 10.1016/j.egypro.2017.03.834
• Languages: English
• ISSNs: 1876-6102

Abstract

Abstract: This study proposed two velocity prediction methods based on updated transition probability matrix (TPM), including Nearest Neighbor Method (NNM) and Fuzzy Set-based Encoding Scheme (FSES). Comparison of mean square errors indicated that the latter offers better prediction accuracy. Reinforcement learning (RL) is applied to select prediction horizon. The results of fuel consumption showed that 7-step prediction horizon is reasonable for model prediction control in the future studies.