The automotive industry is facing a transformation towards the massive digitalization and data-acquisition of the vehicles operation. The exploitation of operational data opens up new opportunities in the energy efficiency improvement of the vehicles. In this regard, the combination of optimization techniques with neural networks and fuzzy systems in one unified framework, known as learning-based energy management strategies, have been identified as promising methods. These learning-based techniques combine the optimized operation with the IF-THEN human-type reasoning simplicity of a fuzzy system through neural-type of learning. Therefore, fuzzy-neural networks are the bridge that allows to learn offline from the optimal operation and design energy management strategy for real time implementation. In this regard, the main contribution of this paper lies on the comparison of a previously developed ANFIS approach with a simpler Neo-Fuzzy neuron based, with the aim to evaluate the tradeoff between accuracy and computational and structural efficiency. The proposed approach represents a fuzzy-neural structure with less parameters for training that is expected to facilitate its future real time application for energy management strategies for each bus from a fleet operating on a predefined route.
|Publication status||Accepted/In press - 2020|
|MoE publication type||Not Eligible|
|Event||IEEE Vehicular Power and Propulsion Conference, IEEE VPPC 2020: Online - Virtual|
Duration: 26 Oct 2020 → 18 Dec 2020
|Conference||IEEE Vehicular Power and Propulsion Conference, IEEE VPPC 2020|
|Abbreviated title||VPPC 2020|
|Period||26/10/20 → 18/12/20|
- neural network
- fuzzy logic
- neo-fuzzy neuron
- dynamic optimization
Lopez-Ibarra, J. A., Gaztanaga, H., Todorov, Y., & Pihlatie, M. (Accepted/In press). Learning Based Energy Management Strategy Offline Trainers Comparison for Plug-In Hybrid Electrical Buses. Paper presented at IEEE Vehicular Power and Propulsion Conference, IEEE VPPC 2020, .