Fast Predictive Eco-Driving Control Of Connected Electric Vehicles

Electrification and informatization have provided unprecedented opportunities for achieving efficient and low-carbon automobiles.In this context,conducting research on eco-driving control of connected electric vehicles(CEVs)have significant strategic implications for China to achieve its "dual carbon" goals.However,how to fully integrate real-time connected information to achieve predictive eco-driving control and improve energy efficiency still faces numerous scientific problems and challenges.In terms of eco-driving optimization problem formulation,effectively handling the spatial-temporal coupling constraints from different vehicle-road hierarchical connected information and fully exploiting energy-saving potential lacks effective theoretical support.Regarding solving eco-driving optimization problems,the introduction of more external connected information has made eco-driving optimization a high-dimensional,complex,nonlinear optimal control problem.How to achieve fast solutions under limited onboard hardware computing capabilities remains a key challenge restricting the application of advanced algorithms in real vehicles.To address these challenges,this paper focuses on theoretical and applied research of predictive eco-driving control based on the availability of connected information for CEVs.The specific research content includes:Firstly,to tackle the challenges of information synchronization and large prediction time-domain spans faced by eco-driving control integrated with road geographic information,a convex-concave optimization problem formulation and a general computational efficient algorithm for distance-domain cruising control are proposed.By exploiting the convex characteristic of the system,the key system dynamic equations undergo lossless convex relaxation without introducing additional simplifications such as model linearization,ensuring both the accuracy and consistency of problem formulation through theoretical proofs.Furthermore,a general algorithm called real-time iteration sequential convex programming is proposed to address the convex-concave model predictive control underlying cruise control,significantly enhancing computational efficiency.Secondly,for eco-driving control considering frequent mode switching of powertrain systems due to preceding vehicle behavior information,an economic predictive cruise control method based on smooth nonlinear programming is proposed.By avoiding model assumptions and simplifications through smooth reconstruction of control variables,the non-smoothness of system dynamic equations caused by motor drive-brake mode switching is addressed.The problem is successfully transformed into a smooth nonlinear programming suitable for numerical solutions,with theoretical proofs ensuring consistency of optimal solutions before and after reconstruction.Furthermore,the balance between energy-efficiency and real-time performance is achieved through realtime iteration sequential quadratic programming algorithm.Next,addressing the issue of integrating traffic light information affected by the switching of traffic light time and fixed spatial constraints,a method for constructing energy-optimal speed planning problems under characteristic time constraints,combined with direct multiple shooting and time-scaling,is proposed.The spatial-temporal constraints are formulated as characteristic time constraints and embedded into a unified optimization problem,overcoming the inherent assumptions and errors introduced by existing parameter optimization frameworks.All decision variables in different temporal and spatial dimensions are transformed into numerical optimization variables,achieving integrated fast solutions for green light time window selection,intersection crossing time,and speed trajectory,significantly reducing energy consumption.Finally,for the challenge in coordinating traffic efficiency and energy efficiency in complex urban road networks,a method for constructing timeenergy optimal control problems under characteristic time constraints is proposed.Based on the proposed real-time iteration sequential algorithm and numerical transformation method,fast solutions to this non-standard dynamic optimization problem are achieved.Validation is conducted based on factors such as the number of vehicles in coordinated optimization,signal light quantities,and signal light timing.Subsequently,a thorough validation of the fast predictive eco-driving control strategy proposed in this paper was performed under diverse conditions such as highway,elevated road,suburban,and urban scenarios.The outcomes demonstrate an overall reduction of approximately 2.83%in energy consumption,accompanied by the achievement of millisecond-level resolution in on-board controller computation.In conclusion,predictive eco-driving control of CEVs constitutes a multidisciplinary interdisciplinary engineering problem.Through optimal control empowerment and numerical optimization techniques,challenges such as the efficient integration of connected information and fast solution of optimal control problem within automotive control have been addressed,thereby promoting the improvement of energy-efficient and low-carbon levels.