Today's drivers of battery electric vehicles must deal with limited driving range in a sparse charging infrastructure. An accurate prediction of energy demand and driving range is therefore important and enables reliable routing and charge planning applications. Predictions of energy demand entail uncertainty, which can be considered directly with the use of probabilistic prediction algorithms. Machine learning algorithms are frequently applied in this context, but data used to train these algorithms are often distributed over a fleet of connected vehicles. Federated learning can be applied in this setting, but predictive uncertainty is typically not considered. We apply an extension of the federated averaging algorithm to learn probabilistic neural networks and linear regression models in a communication-efficient and privacy-preserving manner. We demonstrate the performance advantage of probabilistic prediction models over deterministic prediction models using proper scoring rules. Furthermore, we show that federated learning can improve the standard, driver-individual learning. Using probabilistic predictions, variable safety margins based on destination attainability can be applied, leading to increased effective driving range and reduced travel time.

Limited range and charging infrastructure leads to range anxiety of electric vehicle drivers. Current range estimation algorithms are deemed unreliable and large safety margins are reserved to prevent the risk of stranding. Range estimation in general depends on two factors: current battery energy content and the energy consumption forecast on the route to destination. This work aims at improving the latter by enhancing the forecast with a notion of uncertainty. The prediction algorithm itself learns from driver and traffic data in a training set to generate accurate, driver-individual energy consumption forecasts. Thereby, a central part of the algorithm is the explicit evaluation of the traffic situation by classifying the traffic phases. With the help of this methodology, individual forecasts can be made more precise since they are highly dependent on surrounding traffic. To demonstrate the validity of the algorithms, the performance is evaluated using real test drive data comprising multiple drivers. On the basis of the performance evaluation, both the superiority of stochastic algorithms over deterministic predictions and the improvement of predictive performance by evaluating explicit traffic phases can be shown. Implementing the proposed methodology in modern day electric vehicles could reduce range anxiety and ultimately increase acceptance of electric mobility worldwide.

Probabilistic predictions with machine learning are important in many applications. These are commonly done with Bayesian learning algorithms. However, Bayesian learning methods are computationally expensive in comparison with non-Bayesian methods. Furthermore, the data used to train these algorithms are often distributed over a large group of end devices. Federated learning can be applied in this setting in a communication-efficient and privacy-preserving manner but does not include predictive uncertainty. To represent predictive uncertainty in federated learning, our suggestion is to introduce uncertainty in the aggregation step of the algorithm by treating the set of local weights as a posterior distribution for the weights of the global model. We compare our approach to state-of-the-art Bayesian and non-Bayesian probabilistic learning algorithms. By applying proper scoring rules to evaluate the predictive distributions, we show that our approach can achieve similar performance as the benchmark would achieve in a non-distributed setting.

Society relies on electric mobility to decrease the problems associated with local emissions and global climate change. One key factor for the success of electric vehicles is their everyday usability. Today's users must deal with limited driving range in a sparse charging infrastructure. In this work, we investigate central features influencing the everyday usability of electric vehicles, such as battery capacity, charging infrastructure, range prediction accuracy and vehicle concepts. Since the influence of these aspects depend on each other, they cannot be examined separately. Therefore, we created a stochastic simulation framework with vehicle models and map data to calculate a large amount of different feature variations. One of our key findings is that battery capacities beyond 100 kWh are not feasible. In addition, we stated the importance of an accurate range estimation algorithm and a dense network of high-performance charging points for everyday usability. Taking these results into consideration would help policy makers and automobile manufacturers achieve worldwide acceptance of electric vehicles.

Due to sparse charging infrastructure and short driving ranges, drivers of battery electric vehicles (BEVs) can experience range anxiety, which is the fear of stranding with an empty battery. To help eliminate range anxiety and make BEVs more attractive for customers, accurate range estimation methods need to be developed. In recent years, many publications have suggested machine learning algorithms as a fitting method to achieve accurate range estimations. However, these algorithms use a large amount of data and have high computational requirements. A traditional placement of the software within a vehicle's electronic control unit could lead to high latencies and thus detrimental to user experience. But since modern vehicles are connected to a backend, where software modules can be implemented, high latencies can be prevented with intelligent distribution of the algorithm parts. On the other hand, communication between vehicle and backend can be slow or expensive. In this article, an intelligent deployment of a range estimation software based on ML is analyzed. We model hardware and software to enable performance evaluation in early stages of the development process. Based on simulations, different system architectures and module placements are then analyzed in terms of latency, network usage, energy usage, and cost. We show that a distributed system with cloud‐based module placement reduces the end‐to‐end latency significantly, when compared with a traditional vehicle‐based placement. Furthermore, we show that network usage is significantly reduced. This intelligent system enables the application of complex, but accurate range estimation with low latencies, resulting in an improved user experience, which enhances the practicality and acceptance of BEVs.