{"id":{"repo_id":"colostate","oai_identifier":"oai:mountainscholar.org:10217/181369"},"canonical_url":"https://search.dev.ndltd.org/etd/colostate/oai:mountainscholar.org:10217/181369","repository":{"repo_id":"colostate","name":"Colorado State University","base_url":"https://api.mountainscholar.org/server/oai/request"},"display":{"title":"Predicting hybrid vehicle fuel economy and emissions with neural network models trained with real world data","abstract":"Physics-based hybrid vehicle simulation models for fuel economy (FE) exist but are computationally and financially expensive. These models simulate aspects of real-world drive cycles that include the driving environment, thermal management, driver input, and powertrain component behavior. In this study, an alternative method of hybrid vehicle FE simulation is developed by training and testing a time series neural network (NN) model using real world, on-road data. This enables NN models to model many aspects of on-road vehicle dynamics, like regular traffic stops, turning, and irregular accelerations and stops. Unlike the physics-based models, NNs have the advantage of lower computational costs, which could be utilized in near-real-time vehicle system control to determine optimal velocity planning and powertrain control. Models trained in this study used velocity-time traces as an input to predict instantaneous FE. The NN model predicted fuel economy within a mean absolute error of 0-5% for on-road measurements over a 40 minute, real world, city and highway drive cycle. NN models trained with varying lengths of datasets did not improve with training data longer than 35 minutes. When trained with this method, NN models were accurate when tested with data from multiple days of tests and various drive cycles. Multiple NN models were also trained with hybrid vehicles with varying control system settings. NNs can only successfully model a vehicle whose control system settings reflect the training of the model. These results are expected to improve with more comprehensive drive cycle data that includes data from different elevations and various climatic conditions. The predictions from the FE NN model were compared against predictions from the physics-based Autonomie model and a custom HEV simulation model developed at Colorado State University. NNs outperform these models when tested with on-road data to predict FE of a known vehicle. Using a portable emissions monitoring system (PEMS), NN models were also able to predict nitrous oxides and particulate matter emissions with <5% mean absolute error. The NN model method could be used to improve emissions estimates by capturing differences between real world and laboratory tested emissions. Recording and including more data from the vehicle and devices like the PEMS could further improve these NN models.","abstract_html":"Physics-based hybrid vehicle simulation models for fuel economy (FE) exist but are computationally and financially expensive. These models simulate aspects of real-world drive cycles that include the driving environment, thermal management, driver input, and powertrain component behavior. In this study, an alternative method of hybrid vehicle FE simulation is developed by training and testing a time series neural network (NN) model using real world, on-road data. This enables NN models to model many aspects of on-road vehicle dynamics, like regular traffic stops, turning, and irregular accelerations and stops. Unlike the physics-based models, NNs have the advantage of lower computational costs, which could be utilized in near-real-time vehicle system control to determine optimal velocity planning and powertrain control. Models trained in this study used velocity-time traces as an input to predict instantaneous FE. The NN model predicted fuel economy within a mean absolute error of 0-5% for on-road measurements over a 40 minute, real world, city and highway drive cycle. NN models trained with varying lengths of datasets did not improve with training data longer than 35 minutes. When trained with this method, NN models were accurate when tested with data from multiple days of tests and various drive cycles. Multiple NN models were also trained with hybrid vehicles with varying control system settings. NNs can only successfully model a vehicle whose control system settings reflect the training of the model. These results are expected to improve with more comprehensive drive cycle data that includes data from different elevations and various climatic conditions. The predictions from the FE NN model were compared against predictions from the physics-based Autonomie model and a custom HEV simulation model developed at Colorado State University. NNs outperform these models when tested with on-road data to predict FE of a known vehicle. Using a portable emissions monitoring system (PEMS), NN models were also able to predict nitrous oxides and particulate matter emissions with &lt;5% mean absolute error. The NN model method could be used to improve emissions estimates by capturing differences between real world and laboratory tested emissions. Recording and including more data from the vehicle and devices like the PEMS could further improve these NN models.","abstract_has_math":false,"creators":["Galang, Abril, author","Jathar, Shantanu, advisor","Bradley, Thomas, advisor","Anderson, Chuck, committee member"],"institution":"Colorado State University. Libraries","degree_name":"Master of Science (M.S.)","degree_level":"Masters","degree_discipline":"Mechanical Engineering","degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2017,"date_issued":"2017","date_published":"2017","updated_at":"2026-07-27T19:13:00Z","subjects":["fuel economy","neural network","real world","hybrid electric vehicle","emission","PEMS"],"languages":["eng","English"],"rights":["Copyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright."],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://doi.org/10.25675/3.018452"],"render_values":[{"text":"https://doi.org/10.25675/3.018452","href":"https://doi.org/10.25675/3.018452","code":true}]}]},"links":{"outbound_url":"http://hdl.handle.net/10217/181369","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Galang, Abril, author","Jathar, Shantanu, advisor","Bradley, Thomas, advisor","Anderson, Chuck, committee member"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2017-06-09T15:41:12Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2017-06-09T15:41:12Z"]},{"key":"dc:date.issued","label":"Date","values":["2017"]},{"key":"dc:publisher","label":"Institution","values":["Colorado State University. Libraries"]},{"key":"dc:type","label":"Dc Type","values":["Text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Mechanical Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science (M.S.)"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Colorado State University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["fuel economy","neural network","real world","hybrid electric vehicle","emission","PEMS"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["English"]},{"key":"dc:language.iso","label":"Language (ISO)","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["Galang_colostate_0053N_14096.pdf"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/10217/181369","https://doi.org/10.25675/3.018452"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Physics-based hybrid vehicle simulation models for fuel economy (FE) exist but are computationally and financially expensive. These models simulate aspects of real-world drive cycles that include the driving environment, thermal management, driver input, and powertrain component behavior. In this study, an alternative method of hybrid vehicle FE simulation is developed by training and testing a time series neural network (NN) model using real world, on-road data. This enables NN models to model many aspects of on-road vehicle dynamics, like regular traffic stops, turning, and irregular accelerations and stops. Unlike the physics-based models, NNs have the advantage of lower computational costs, which could be utilized in near-real-time vehicle system control to determine optimal velocity planning and powertrain control. Models trained in this study used velocity-time traces as an input to predict instantaneous FE. The NN model predicted fuel economy within a mean absolute error of 0-5% for on-road measurements over a 40 minute, real world, city and highway drive cycle. NN models trained with varying lengths of datasets did not improve with training data longer than 35 minutes. When trained with this method, NN models were accurate when tested with data from multiple days of tests and various drive cycles. Multiple NN models were also trained with hybrid vehicles with varying control system settings. NNs can only successfully model a vehicle whose control system settings reflect the training of the model. These results are expected to improve with more comprehensive drive cycle data that includes data from different elevations and various climatic conditions. The predictions from the FE NN model were compared against predictions from the physics-based Autonomie model and a custom HEV simulation model developed at Colorado State University. NNs outperform these models when tested with on-road data to predict FE of a known vehicle. Using a portable emissions monitoring system (PEMS), NN models were also able to predict nitrous oxides and particulate matter emissions with <5% mean absolute error. The NN model method could be used to improve emissions estimates by capturing differences between real world and laboratory tested emissions. Recording and including more data from the vehicle and devices like the PEMS could further improve these NN models."]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["born digital","masters theses"]},{"key":"dc:title","label":"Title","values":["Predicting hybrid vehicle fuel economy and emissions with neural network models trained with real world data"]}]}],"canonical_facts":{"dc:creator":["Galang, Abril, author","Jathar, Shantanu, advisor","Bradley, Thomas, advisor","Anderson, Chuck, committee member"],"dc:date.accessioned":["2017-06-09T15:41:12Z"],"dc:date.available":["2017-06-09T15:41:12Z"],"dc:date.issued":["2017"],"dc:description.abstract":["Physics-based hybrid vehicle simulation models for fuel economy (FE) exist but are computationally and financially expensive. These models simulate aspects of real-world drive cycles that include the driving environment, thermal management, driver input, and powertrain component behavior. In this study, an alternative method of hybrid vehicle FE simulation is developed by training and testing a time series neural network (NN) model using real world, on-road data. This enables NN models to model many aspects of on-road vehicle dynamics, like regular traffic stops, turning, and irregular accelerations and stops. Unlike the physics-based models, NNs have the advantage of lower computational costs, which could be utilized in near-real-time vehicle system control to determine optimal velocity planning and powertrain control. Models trained in this study used velocity-time traces as an input to predict instantaneous FE. The NN model predicted fuel economy within a mean absolute error of 0-5% for on-road measurements over a 40 minute, real world, city and highway drive cycle. NN models trained with varying lengths of datasets did not improve with training data longer than 35 minutes. When trained with this method, NN models were accurate when tested with data from multiple days of tests and various drive cycles. Multiple NN models were also trained with hybrid vehicles with varying control system settings. NNs can only successfully model a vehicle whose control system settings reflect the training of the model. These results are expected to improve with more comprehensive drive cycle data that includes data from different elevations and various climatic conditions. The predictions from the FE NN model were compared against predictions from the physics-based Autonomie model and a custom HEV simulation model developed at Colorado State University. NNs outperform these models when tested with on-road data to predict FE of a known vehicle. Using a portable emissions monitoring system (PEMS), NN models were also able to predict nitrous oxides and particulate matter emissions with <5% mean absolute error. The NN model method could be used to improve emissions estimates by capturing differences between real world and laboratory tested emissions. Recording and including more data from the vehicle and devices like the PEMS could further improve these NN models."],"dc:format.medium":["born digital","masters theses"],"dc:identifier":["Galang_colostate_0053N_14096.pdf"],"dc:identifier.uri":["http://hdl.handle.net/10217/181369","https://doi.org/10.25675/3.018452"],"dc:language":["English"],"dc:language.iso":["eng"],"dc:publisher":["Colorado State University. Libraries"],"dc:rights":["Copyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright."],"dc:subject":["fuel economy","neural network","real world","hybrid electric vehicle","emission","PEMS"],"dc:title":["Predicting hybrid vehicle fuel economy and emissions with neural network models trained with real world data"],"dc:type":["Text"],"thesis:degree_discipline":["Mechanical Engineering"],"thesis:degree_level":["Masters"],"thesis:degree_name":["Master of Science (M.S.)"],"thesis:institution_name":["Colorado State University"]},"updated_at":"2026-07-27T19:13:00Z"}