In this study, a novel Power Management System (PMS) is proposed for Hybrid Electric Vehicles/Vessels (HEV) having cyclic operational profiles. In these vehicles, the PMS is in charge for meeting the demand electrical power by means of a set of internal combustion engines and an Energy Storage System (ESS). This leads PMS requires an advanced intelligent computer-aided tool to optimize the performance of the generation and transmission of the power among different power sources. The PMS in this study consists of offline and online control layers. An Energy Management System (EMS) is introduced as the first layer which is inherently composed of three main sections, learning package, data processor, and optimizer. Unsupervised exclusive machine learning algorithms are employed to estimate the previously unseen operational profile for the vehicle on the basis of all past loading demand profile. $k-means$ and $k-medoids$ are the two learning algorithms are adapted to the proposed control system. The simulation results represent a high efficiency of these two algorithms with respect to an ideal controller in which the future is known for the controller. To reduce the dimensionality of the post-learned data, segmentation as a data mining technique is then used. Adaptive Piece-wise Constant Approximation algorithm is applied to do so. To synchronize and supervise the distribution of the load profile between engines and ESS, post-segmentation profile is then dispatched to the EMS optimizer which is adopting Mixed-Integer Linear Programming as a mathematical programming technique. This layer in general is in charge for optimizing the State of Charge (SOC) in the ESS for the next cycle of operation. An online controller is then employed in the second layer to controller. It controls real-time SOC in ESS. In this study, two Proportional Integral Derivative (PID) controllers are used in the online control system which are monitoring the SOC of the ESS and engines rotational speed. The PMS has been implemented in the Simulink environment of MATLAB and the results corresponding to a case study of a different configurations of Hybrid-Electric Ferry (HEF) are presented. The results are dramatically showing the reduction of fuel consumption up to 25\% and reliability of the control system for wide range of variation in the operation.
In questo studio, un nuovo sistema di gestione della potenza (PMS) viene proposto per veicoli / veicoli elettrici ibridi (HEV) con profili operativi ciclici. In questi veicoli, il PMS è incaricato di soddisfare la domanda di energia elettrica mediante una serie di motori a combustione interna e un sistema di stoccaggio dell'energia (ESS). Questo porta PMS richiede uno strumento avanzato computer intelligente per ottimizzare le prestazioni della generazione e trasmissione della potenza tra diverse fonti di energia. Il PMS in questo studio è costituito da livelli di controllo offline e online. Un sistema di gestione dell'energia (EMS) viene introdotto come primo livello che è intrinsecamente composto da tre sezioni principali, un pacchetto di apprendimento, un processore di dati e un ottimizzatore. Gli algoritmi di apprendimento automatico senza supervisione sono utilizzati per stimare il profilo operativo precedentemente non visto per il veicolo sulla base di tutti i precedenti profili di domanda di carico. $ k-significa $ e $ k-medoidi $ sono i due algoritmi di apprendimento adattati al sistema di controllo proposto. I risultati della simulazione rappresentano un'alta efficienza di questi due algoritmi rispetto a un controller ideale in cui il futuro è noto per il controller. Per ridurre la dimensionalità dei dati post-apprendimento, viene utilizzata la segmentazione come tecnica di data mining. Algoritmo Adaptive Piece Approximation Approximation viene applicato per farlo. Per sincronizzare e supervisionare la distribuzione del profilo di carico tra motori ed ESS, il profilo di post-segmentazione viene quindi inviato all'ottizzatore EMS che sta adottando la programmazione lineare a numero intero come tecnica di programmazione matematica. Questo strato in generale è responsabile dell'ottimizzazione dello stato di carica (SOC) nell'ESS per il prossimo ciclo di operazioni. Un controller online viene quindi impiegato nel secondo livello del controller. Controlla SOC in tempo reale in ESS. In questo studio, due controller proporzionali di derivazione integrale (PID) sono utilizzati nel sistema di controllo online che sta monitorando il SOC dell'ESS e la velocità di rotazione dei motori. Il PMS è stato implementato nell'ambiente Simulink di MATLAB e vengono presentati i risultati corrispondenti a un caso di studio di diverse configurazioni di Hybrid-Electric Ferry (HEF). I risultati mostrano drammaticamente la riduzione del consumo di carburante fino al 25% e l'affidabilità del sistema di controllo per un'ampia gamma di variazioni nell'operazione.
Application of Machine Learning and Mathematical Programming in Optimization of Control System in Hybrid-Electric Vehicles/Vessels Having Cyclic Operations
MOHAMMADZADEH, NAVID
2017/2018
Abstract
In this study, a novel Power Management System (PMS) is proposed for Hybrid Electric Vehicles/Vessels (HEV) having cyclic operational profiles. In these vehicles, the PMS is in charge for meeting the demand electrical power by means of a set of internal combustion engines and an Energy Storage System (ESS). This leads PMS requires an advanced intelligent computer-aided tool to optimize the performance of the generation and transmission of the power among different power sources. The PMS in this study consists of offline and online control layers. An Energy Management System (EMS) is introduced as the first layer which is inherently composed of three main sections, learning package, data processor, and optimizer. Unsupervised exclusive machine learning algorithms are employed to estimate the previously unseen operational profile for the vehicle on the basis of all past loading demand profile. $k-means$ and $k-medoids$ are the two learning algorithms are adapted to the proposed control system. The simulation results represent a high efficiency of these two algorithms with respect to an ideal controller in which the future is known for the controller. To reduce the dimensionality of the post-learned data, segmentation as a data mining technique is then used. Adaptive Piece-wise Constant Approximation algorithm is applied to do so. To synchronize and supervise the distribution of the load profile between engines and ESS, post-segmentation profile is then dispatched to the EMS optimizer which is adopting Mixed-Integer Linear Programming as a mathematical programming technique. This layer in general is in charge for optimizing the State of Charge (SOC) in the ESS for the next cycle of operation. An online controller is then employed in the second layer to controller. It controls real-time SOC in ESS. In this study, two Proportional Integral Derivative (PID) controllers are used in the online control system which are monitoring the SOC of the ESS and engines rotational speed. The PMS has been implemented in the Simulink environment of MATLAB and the results corresponding to a case study of a different configurations of Hybrid-Electric Ferry (HEF) are presented. The results are dramatically showing the reduction of fuel consumption up to 25\% and reliability of the control system for wide range of variation in the operation.| File | Dimensione | Formato | |
|---|---|---|---|
|
NavidMohammadzadehMasterThesisMechanicalEngineering.pdf
accessibile in internet per tutti
Descrizione: Thesis Text
Dimensione
7.5 MB
Formato
Adobe PDF
|
7.5 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/244097