A novel cascade approach to control variables optimisation for advanced series-parallel hybrid electric vehicle power-train

• Published in: Applied energy 2020-10, Vol.276, p.115488, Article 115488
• Main Authors: Cipek, Mihael, Kasać, Josip, Pavković, Danijel, Zorc, Davor
• Format: Article
• Language: English
• Subjects: Control variable optimisation; Dynamic programming; Fuel economy improvement; Gradient methods; Hybrid electric vehicle
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2020.115488

•A novel cascade optimisation algorithm has been proposed.•Optimisation is performed over a simplified (backward) powertrain model.•Results of cascade optimisation are compared to dynamic programming benchmark.•Accuracy is improved by increasing the gradient algorithm’s numerical precision.•Cascade optimisation convergence rate is eight time faster than dynamic programming. In recent years, road vehicles are being increasingly equipped with hybrid electric power-trains in order to provide significant gains in fuel economy and reductions in greenhouse gases emissions. Since hybrid power-trains consist of two or more different energy sources, many of their variants are present nowadays which leads to many open questions in terms of hybrid electric power-train structure selection, components sizing and energy management control, which all have influence on the power-train purchase cost and efficiency. The control variables optimisation is crucial in order to find the set of optimal control rules for different power-train operating regimes which would yield the minimum possible fuel consumption. Among different control variable optimisation methods, the dynamic programming approach is usually used in literature, because of its unique feature to find the global optimum solution with a certain degree of precision. However, this optimisation method also requires significant computing power and its application is limited to low-order systems. Having this in mind, this paper evaluates the benefits of innovative cascade approach to hybrid electric vehicle control variable optimisation wherein dynamic programming is combined with a gradient-based optimisation algorithm in a systematic and a straightforward way in order to significantly reduce the optimisation execution time and also to increase the precision of the globally-optimal result.