Prediction of Electric Buses Energy Consumption from Trip Parameters Using Deep Learning

Abstract

The energy demand of electric buses (EBs) is a very important parameter that should be considered by transport companies when introducing electric buses into the urban bus fleet. This article proposes a novel deep-learning-based model for predicting energy consumption of an electric bus traveling in an urban area. The model addresses two important issues: accuracy and cost of prediction. The aim of the research was to develop the deep-learning-based prediction model, which requires only the data readily available to bus fleet operators, such as location of the bus stops (coordinates, altitude), route traveled, schedule, travel time between stops, and to find the most suitable type and configuration of neural network to evaluate the model. The developed prediction model was assessed with different types of deep neural networks using real data collected for several bus lines in a medium-sized city in Poland. Conducted research has shown that the deep learning network with autoencoders (DLNA) neural network allows for the most accurate energy consumption estimation of 93%. The proposed model can be used by public transport companies to plan driving schedules and energy management when introducing electric buses.

Data Driven Energy Economy Prediction for Electric City Buses Using Machine Learning

“…This finding significantly reduces the number of variables to be measured on the one hand and on the other it animates researchers in the field to explore the feature space and selection methods more deeply. Like other authors, see [21], [23], [29], [56], we have also compared and discussed our models regarding their overall applicability. Operators can clearly benefit from our most robust and accurate model based on NCAFS and GPR or they could choose the reduced bagged RF combined with oobImp, as it is less complex and could be easily implemented with a lazy learning algorithm on a vehicle platform with low computational capabilities.…”

“…city buses, buttressing the motivation of our work. Pamula et al [23] used both deep learning and classical neural networks to forecast the energy demand of electric buses. These prediction models utilized actual data obtained from various bus lines.…”

Analysis on the Bus Arrival Time Prediction Model for Human-Centric Services Using Data Mining Techniques

“…The support vector machine algorithm finds the equation for the optimal hyperplane from the training data and uses it for later predictions. The confidence value for the classifications will be directly proportional to the distance between the current data point and the hyperplane, which is the decision boundary [ 27 ]. The main purpose behind finding this optimal hyperplane that is far from all data points is to maximize the confidence value for future predictions.…”

“…e support vector machine algorithm finds the equation for the optimal hyperplane from the training data and uses it for later predictions. e confidence value for the classifications will be directly proportional to the distance between the current data point and the hyperplane, which is the decision boundary [27].…”

Battery Electric Vehicles: How Many Gears? A Technical–Economic Analysis

“…Battery Electric Vehicles (BEVs) are considered the most promising solution to decarbonize road transport in the future, especially for light vehicles: other solutions such as e-fuels and hydrogen, are usually considered less effective, and possibly confined to heavyduty road transport, as well as ships and aircraft [1][2][3]. The natural torque characteristic of BEVs makes it possible for them to use a single-speed mechanical transmission, as electric drives can deliver the maximum torque starting from zero speed and provide maximal power across a wide operating range.…”