An improved vehicle to the grid method with battery longevity management in a microgrid application

Highlights

• Battery lifetime degradation is considered in the proposed V2G scheme.

• The proposed V2G method achieves frequency response in the microgrid.

• The prediction of the V2G capacity is conducted in the V2G scheduling.

• The method is verified in a MG scenario with reduction of charge/discharge cycles.

Abstract

This paper proposed an improved vehicle-to-grid (V2G) scheduling approach for the frequency control with the advantage of protecting the batteries hence saving the battery lifetime during grid connected service. The proposed methodology is improved in two ways. Firstly, to give a prediction of the available electric vehicle (EV) battery capacity in the control time-step for the V2G service, a deep learning based prediction is developed. Secondly, this study advances the previous V2G method by adding the quantitative analysis of the battery cycle life into the V2G optimization. The accurate prediction of the schedulable battery capacity based on the LSTM algorithm is shown very effective in the power system frequency control. Also, compared with the previous method that without battery lifetime control, the proposed method benefits in the reduction of charge/discharge cycles.

Introduction

The increasingly share of renewable sources integrated to the network, strict set for the reduction of greenhouse gas emissions, and the need for providing clean energy, call for a paradigm shift in energy systems. The efficient power generations and energy consumptions are playing the key factors in this transformation [[1], [2], [3]]. On the one hand, the changes are obvious in the power gird as the generations are moving from the large centralized power plants to the distributed renewable sources [4,5]. The transport electrification, on the other hand, is playing a vital role in this transformation has been recognized by industry and policy makers [[6], [7], [8]]. The ambitious targets are published to promote the transport evolution by many countries [9,10]. For example, the United Kingdom government has announced a ban on the sale of the traditional diesel and petrol cars and vans after 2040 [11]. Inherently, the EVs or more specifically, the power batteries are regarded as the intruders for the traditional power grid, and with the large-scale adoption of electric vehicles, their uncoordinated charge demands are adding strains on the grid infrastructure [12]. As a result, if electric vehicle charging is left uncoordinated, the adoption of electric vehicles is expected to cause significant system power fluctuations, which will bring significant challenges on both system frequency and voltage stability reported by many researches [13,14].

Nevertheless, the EV power batteries is regarded as the “moving energy storage” that offer the means to enhance power system flexibility especially for the grid and achieve uninterrupted operation by deferring their demand in time and even space and acting as dynamic storage devices. Therefore, it comes to the concept of vehicle-to-grid that effectively integrates the aggregated EVs into the microgrid as distributed energy resources to act as controllable generations or loads achieving the benefit of frequency regulating, voltage control, techno-economic operating, etc. [[15], [16], [17], [18], [19]]. The increasing penetration of renewable generation, the updated advances in energy storages and the substantial uptake of electrification of transport itself, is incessantly imposing unprecedented complexity and uncertainty on the V2G scheduling. With the increasing penetration of renewable energy resources, the development of high-performance V2G scheduling strategies has attracted much attention in global academic and industry communities [[23], [24], [25]]. Electric vehicles could provide ancillary services for the grid, but to enable this benefit, a key issue that should be addressed first is how to predict the V2G schedulable capacity information to meet different utility demands of power dispatch.

The statistical forecasting is widely used to make the capacity prediction based on historical data [26,27]. Ref [28] interduces a power management method for integrating the EVs to the gird with fuzzy logic algorithm achieving an excellent operational resource scheduling. In the V2G scheme, the conventional scheduling could hardly address the emerging opportunities regarding to increased system information and complexity. The V2G control need to deal with not only the regular charge behavior under prediction scheme but the short-term uncertainty as well. Deep learning algorithms have been investigated to be used different applications such as fault detection [29,30], demand side forecasting in power systems [31] and traffic prediction in transportation system [32]. The long short-term memory neural network is good at mining deep structure features in time-series data [33,34], hence used for the battery capacity prediction.

To mitigate the battery degradation problem, on the one hand, the new V2G method should functionally take the battery degradation into account, and on the other hand, the new V2G scheme should provide the evidences with quantitative analysis of the techno-economic advantages to the customers to encourage their participation. The trade-off between the V2G service and the battery lifetime degradation is very difficult to reach. In addition, the objective function is usually not simple linear or quadratic, so the regular convex optimization method is not suitable in this case [35]. The introducing of a quantitative index of battery degradation makes it much worse that the objective function is non-gradient, which fails the regular gradient descent algorithms [36] and the non-gradient optimization is normally used to solve this kind of problem. Different methods of none-gradient optimization can be found with different characteristics in the knowledge field in different kind of applications [[37], [38], [39], [40], [41]]. Liu et al. provide an good example by developing a multi-objective optimization strategy to optimize to maximize the fundamental frequency as well as minimize the dynamic displacement simultaneously [42]. The particle swarm optimization (PSO) has been investigated to be used in different applications [43,44]. Huo et al. presents an decomposed hybrid particle swarm method to achieve the optimal operation of interconnected energy hubs [44]. For the multi-objective optimization problem, the study presented by Ref. [45] developed a PEV charging coordination method based on fuzzy discrete particle swarm optimization, where several optimization objectives are combined based on fuzzy logic. However, the previous approaches did not consider the battery aging process and the quantitative analysis of the battery aging effect during the V2G services. To solve this problem, this study developed an PSO algorithms combined with the rain-flow cycle counting to reflect the battery aging process in the V2G scheduling. Comparing to the empirical-based or data-driven battery life estimation models, it is easier to quantify the cycles in rain-flow counting algorithm-based battery life estimation model [46]. Therefore, the EV battery charge/discharge cycle number is used in the proposed PSO algorithm.

The proposed EV battery available capacity prediction method and V2G battery anti-aging scheduling approach is verified to be effective by in the power system frequency regulation service. Compared with the previous method that without battery lifetime degradation consideration, the proposed method benefits in the reduction in charge/discharge cycles.