Research On Electric Vehicle Logistics Scheduling Problem Based On Evolutionary Algorithms

Electric vehicles have gradually become an important new energy transport tool in green logistics scheduling,because of its obvious advantages of clean and energy saving.Different from the traditional vehicle scheduling problem based on fuel vehicle,the logistics scheduling problem based on electric vehicle(EV)should consider both capacity constraints and electricity constraints.As a result,the feasible region of the solution space is smaller and it’s harder to get solution.Although heuristic algorithms are widely used to solve the logistics scheduling of electric vehicles,they are easy to fall into local optimum because of the single-point search method.So it’s not suitable for solving multi-objective optimization problems.Evolutionary algorithms(EA)are a kind of population-based swarm intelligence optimization algorithms.They have a good global search characteristic and are widely used to solve the traditional vehicle routing problem and facility location problem.However,there are few studies on developing the EA to solve the electric vehicle logistics scheduling problem,and the search efficiency needs to be improved because the evolutionary algorithm adopts the population-based search method.Therefore,further research about the electric vehicle logistics scheduling problem based on evolutionary algorithm is carries out in this paper,including two hot issues: the electric vehicle routing problem(EVRP)and the electric vehicle location routing problem(ELRP).The main research work of this paper is as follows:(1)For the electric vehicle routing problem(EVRP),a co-evolutionary algorithm based on two populations(COEA)is proposed.EVRP needs to consider both capacity constraints and electricity constraints,which leads to slow search speed and the solution is difficulty.Therefore,the proposed algorithm adopts the idea of two-population coevolution,and uses simple vehicle routing problem with capacity constraints(CVRP)to assist solving the complex EVRP.Firstly,in order to facilitate the knowledge transfer among these two heterogeneous problems,a distance adjacency matrix is designed to represent excellent solutions in CVRP and EVRP,which contains richer routing information and is convenient for later learning model to extract problem domain knowledge.Then,the Autoencoding is used to construct the relationship between these two problems through the distance adjacency matrix,which represents the transformation relationship under the constraints of different scenarios with the same distribution of customers and facility points.Finally,the solutions transformed from CVRP coevolves with the EVRP population.By learning the customers sequence of CVRP,the route of EVRP is promoted and the search efficiency of the algorithm is improved.The proposed algorithm is verified on large scale datasets with five state-ofthe-art algorithms.The experimental results show that the COEA has a good performance on the large-scale electric vehicle routing problem,and the quality of the solutions obtained has a clear advantage compared with other five algorithms.(2)An Accelerating two-phase evolutionary algorithm(A-TPEA)is proposed for the electric vehicle location-routing problem(ELRP).Due to the ELRP should consider the location of charging facilities while planning the distribution route,it’s more difficult for the solution approaches.The proposed algorithm adopts the framework of location-routing two-phase iterative optimization,which is commonly used by heuristic algorithms at present.The algorithm framework is adopted and introduces interpolation method and the surrogate model in each stage,and generates high-quality location and routing offspring respectively to accelerate the solving efficiency of EA to this problem.Specifically,in the electeic vehicle routing optimization phase,an interpolation method is developed to extract the frequent visiting orders existing in the historical elite solutions,and these visiting orders are used to generate potential routing offspring that can accelerate convergence to the optimal regions.In the charging facility location optimization phase,the surrogate model is used to construct the corresponding relationship between the route and location,which can directly output a promising location scheme for the given routing offspring and thus improve the search efficiency.In order to verify the effectiveness of the proposed algorithm,it is compared with other five state-of-the-art algorithms on the public test sets.The experimental results show that A-TPEA has better performance on different scale instances and the proposed algorithm’s objective functions are basically better than other five algorithms.