As the adoption of battery electric buses (BEBs) in public transportation systems grows, the need for precise energy consumption forecasting becomes increasingly important. Accurate predictions are essential for optimizing routes, charging schedules, and ensuring adequate operational range. This paper introduces an innovative forecasting methodology that combines a propulsion and auxiliary energy model with a novel concept, the environment generator. This approach addresses the primary challenge in electric bus energy forecasting: estimating future environmental conditions, such as weather, passenger load, and traffic patterns, which significantly impact energy demand. The environment generator plays a crucial role by providing the energy models with realistic input data. This study validates various models with different levels of model complexity against real-world operational data from a case study of over one year with 16 electric buses in Göttingen, Germany. Our analysis thoroughly examines influencing factors on energy consumption, like altitude, temperature, passenger load, and driving patterns. In order to comprehensively understand energy demands under varying operational conditions, the methodology integrates data-driven models and physical simulations into a modular and highly accurate energy predictor. The results demonstrate the effectiveness of our approach in providing more accurate energy consumption forecasts, which is essential for efficient electric bus fleet management. This research contributes to the growing body of knowledge in electric vehicle energy prediction and offers practical insights for transit authorities and operators in optimizing electric bus operations.
The electrification of urban bus fleets is a challenging task, especially for smaller public transport operators. The main challenge lies in the uncertainty about many technical aspects, like range of vehicles under different circumstances or charging times, that are new for the operators. The purpose of this research is to introduce an approach to solve this problem by incorporating all available data from an existing bus fleet and finding an optimal solution with discrete mathematical optimization. Extensive data logging in the project enabled us to leverage tracking data from the whole bus network including trajectories, powertrain data, and operational data. This enabled us to validate assumptions about the energy demand, waiting times, and different traffic situations during the day. To get better insights into the requirements of an urban bus fleet, we simulated the potential electric buses in detail and extracted other necessary data like actual dwell times. Based on the simulation results and processed data, we implemented a linear programming model to search for a cost-optimal configuration of vehicles and charging infrastructure. We tested the framework with a scenario in which we analyzed the solutions with different numbers of diesel buses in the fleet. The application of our algorithm shows that it can produce optimal results in a short amount of time, for a medium-sized city in Germany. We also demonstrate that the flexible and constraint-based formulation of this approach allows it to be incorporated in the planning process of most public transport operators.
The response time of Artificial Neural Network (ANN) inference is critical in embedded systems processing sensor data close to the source. This is particularly important in applications such as predictive maintenance, which rely on timely state change predictions. This study enables estimation of model response times based on the underlying platform, highlighting the importance of benchmarking generic ANN applications on edge devices. We analyze the impact of network parameters, activation functions, and single- versus multi-threading on response times. Additionally, potential hardware-related influences, such as clock rate variances, are discussed. The results underline the complexity of task partitioning and scheduling strategies, stressing the need for precise parameter coordination to optimise performance across platforms. This study shows that cutting-edge frameworks do not necessarily perform the required operations automatically for all configurations, which may negatively impact performance. This paper further investigates the influence of network structure on model calibration, quantified using the Expected Calibration Error (ECE), and the limits of potential optimisation opportunities. It also examines the effects of model conversion to Tensorflow Lite (TFLite), highlighting the necessity of considering both performance and calibration when deploying models on embedded systems.