Sa stanovišta elektroenergetske mreže, električna vozila (EV) predstavljaju prostorno-distribuirano baterijsko skladište električne energije, koje se u okviru naprednog punjenja EV može iskoristiti za razne primjene poput poravnanja opterećenja mreže. Kako bi se ispitala isplativost elektrifikacije ciljane flote vozila i njen utjecaj na električnu mrežu, potrebno je provesti optimiranje punjenja odgovarajuće flote EV primjenom prikladnih optimizacijskih postupaka temeljenih na preciznim i računalno-učinkovitim modelima flote EV, a što predstavlja glavni cilj ovog rada. U radu je predložen novi agregatni model flote EV koji je validiran u odnosu na precizniji, ali i računalno neučinkovitiji, distribuirani model. Pritom je optimiranje punjenja, za svaki od modela, provedeno korištenjem algoritma dinamičkog programiranja (DP), koji za opći nelinearni optimizacijski problem garantira globalno optimalne rezultate. Cilj optimiranja je minimizacija cijene električne energije koja se koristi za punjenje flote EV uz zadovoljenje ograničenja vezanih uz stanje napunjenosti (SoC) baterije te snagu punjenja. Osim jednorazinskog DP optimiranja punjenja, predloženo je i dvorazinsko optimiranje gdje se jednorazinsko DP optimiranje proširuje nadređenom razinom,na kojoj se korištenjem genetskog algoritma dodatno optimira maksimalna snaga punjača,te vrijednost SoC-a na početku svakog voznog ciklusa s ciljem minimizacije operativnih troškova flote EV. U svrhu parametriranja i validiranja predloženog agregatnog modela flote EV, postavljenje model električnog vozila s proširenim dometom (EREV) te je predložena pripadna upravljačka strategija pogona. Upravljačka strategija EREV pogona temelji se na regulatoru zasnovanom na bazi pravila (RB),koji se kombinira sa strategijom minimizacije ekvivalentne potrošnje goriva (ECMS). Pritom se održivost SoC-a baterije postiže SoC regulatorom koji je sadržan u RB regulatoru, dok se ECMS-om postiže optimalnost radnih točaka motora s unutarnjim izgaranjem. Predložena upravljačka strategija ispitana je simulacijski, te usporedbom rezultata simulacija s globalno optimalnim rezultatima dobivenima DP optimiranjem upravljačkih varijabli pogona za raznevozne cikluse. Također su pribavljeni i analizirani podaci o voznim ciklusima konkretne flote dostavnih vozila i o energetskom sustavu pripadnog distributivnog centra. Za potrebe parametriranja modela flote, provedene su simulacije razvijenog EREV modela preko snimljenih voznih ciklusa te preko sintetičkih voznih ciklusa koji statistički vjerno reprezentiraju polazni skup snimljenih voznih ciklusa. Konačno, na temelju razvijenih alata provedena je pilot studija vezana uz ispitivanje mogućnosti elektrifikacije spomenute flote dostavnih vozila za različite scenarije udjela obnovljivih izvora energije i vremenske razdiobe cijene električne energije.
|Sažetak (engleski)|| |
Electric vehicles (EV) are recognized as a technology which will gradually replace the existing conventional vehicles propelled by internal combustion engines. This is important from the standpoint of reducing the driving-related energy consumption and cutting the emissions of greenhouse gases and other pollutants. Furthermore,EVs containing an ample energy storage (usually electrochemical batteries), represent a spatially-distributed energy storage convenient for providing various ancillary services to electric grid and supporting proliferation of renewable energy sources (RES). In order to thoroughly investigate benefits of replacing the conventional vehicles with the electric ones (including energy planning studies), it is crucial to develop precise transport-energy models and use different optimisation techniques to reveal optimal structure, parameters, and management of a considered transport-energy system. In order to facilitate optimisations of integrated transport-energy systems including EV fleet charging management strategies, individual EVs are modelled in this thesis by means of a single aggregate battery, which is validated against a more realistic, although more complex, distributed EV fleet model. The aggregate battery model is parameterised based on several EV fleet-related input time distributions, such as averagestate-of-charge (SoC) time distributions of EVs connecting to and disconnecting from the electric grid, which are obtained by means of simulations of a developed individual EV model including its control strategy. The EV fleet battery charging can generally be managed based on the use of various optimisation algorithms, such as dynamic programming (DP) algorithm, linear or nonlinear programming, mixed integer linear programming, stochastic dynamic programming and heuristic optimisation algorithms. The DP-based charging management optimisation is proposed herein due to its inherent feature of resulting in globally optimal results for the general problem of nonlinear system model with nonlinear constraints. Finally, the developed EV fleet models and related charging management techniques are used to conduct a case study of electrification of a particular delivery vehicle fleet for different levels of local RES penetration. The main aim of the thesis is to establish a systematic approach to modelling and optimal control of transport-energy system with EVs and RES, which includes development of models and optimal control strategies for individual EVs of different types, as well as models and charging management strategies for EV fleets. The thesis is organised inseven chapters, whose content is summarised in what follows. Chapter 1“Introduction” outlines the motivation for the conducted research, presents the literature overview, and provides the main hypothesis and an overview of the thesis. Chapter 2“Modelling and control of an extended range electric vehicle”(EREV) describes the quasi-static modelling of a complex series-parallel EREV configuration and proposes a practical and nearly-optimal control strategy for its powertrain. The control strategy relies on a rule-based (RB) controller which is smoothly combined with an optimal equivalent consumption minimisation strategy (ECMS), where the RB controller ensures sustainable battery charge, while the ECMS ensures optimal selection of the powertrain operating point. In addition, DP-based control variable optimisation is conducted for several characteristic driving cycles in order to gain an insight into the powertrain optimal behaviour and facilitate the feedback control strategy design. The DP optimisation results also serve as a benchmark for the control strategy verification. Chapter 3“Analysis and synthesis of driving cycles” deals with experimental characterisation and analysis of delivery vehicle fleet system of a regional retail company. The vehicle datahave been collected for a fleet of ten delivery vehicles running continuously over a three month period. Next, a detailed statistical analysis of the collected data is presented, in order to provide a basis for future investigation of possible benefits ofreplacing the conventional vehicle fleet with a hypothetical one based on electric vehicles. Finally, the recorded large set of driving cycles is used for the purpose of Markov chain-based synthesis and validation of a small number of representative driving cycles. Such naturalistic driving cycles are used in Chapter 5 for transport energy demand modelling as a part of optimal EV fleet charging management strategy. Chapter 4“Electric vehicle fleet modelling” proposes an aggregate battery modelling approach for an electric vehicle (EV) fleet, which is aimed for energy planning studies of EV-grid integration, as well as for the design of a EV fleet charging management strategy. The proposed approach improves on the existing, basic aggregate battery modellingapproach by accounting for a variable structure of the aggregate battery system (due to vehicle connection/disconnection), including variable state of charge (SoC) constraints and specific time-varying input time-distributions such as those related to average vehicle SoC when arriving to the distribution centre and number of arriving and departing vehicles. The model input distributions are reconstructed from a large set of data related to delivery vehicle fleet driving missions, as well as from results ofsimulation of individual EVs over the full set of recorded driving cycles. The aggregate charging power input is obtained by using a DP-based optimisation algorithm aimed at finding a global optimum in terms of minimal charging electricity cost. For the purpose of proposed model validation and its comparison with the basic model, a distributed fleet vehicle model is developed and accompanied with a computationally efficient, heuristic algorithm for distributing the optimised aggregate charging power input to charging inputs of individual EVs. Chapter 5“Electric vehicle fleet charging optimisation” proposes a DP-based optimisation method of charging an EV fleet described by the aggregate battery model. The main advantage of the proposed approach is that it provides a globally optimal solution, with a relatively non-excessive computational load owing to a low order of the aggregate battery model. In the case of distributed (agent-based) EV fleet model, DP is conducted separately for each EV in successive manner to provide an acceptable computational load, which results in nearly-optimal results in the general case of charging power constraints. The DP charging optimisation approaches are illustrated through a case study of electrified delivery vehicle transport system charged both from the grid and local RES (solar panels). Two scenarios of energy production from RES (with and without excess in RES production), along with several electricity price models are studied. In the case of distributed fleet model, the DP optimisations are conducted and analysed for different levels of maximum charging power which can be taken from a grid. In the case of aggregate battery model, the DP optimisation results are compared with the results obtained by an existing heuristic charging algorithm used in EnergyPLAN software to illustrate the DP algorithm advantages in minimising the charging energy cost and satisfying the aggregate battery charge sustaining conditions. In addition, the single-level DP optimisation of aggregate battery charging power is extended to a bi-level optimisation, where SoC-at-departure and the maximum charging power of an individual EV are optimised at the supervisory level by using a genetic algorithm, with the aim to minimise the operational cost for different charger power levels. Chapter 6“Analysis of techno-economic aspects of delivery vehicle fleet electrification” integrates the methods and numerical tools developed in the previous chapters through a pilot study of designing and analysing a transport-energy system containing a delivery EV fleet and production from local RES. The considered EV fleet is based on series-configuration EREV trucks of comparable speed and torque characteristics as the existing conventional trucks. The overall transport-energy system model also includes prediction of hourly consumption of electricity within the distribution centre and RES production. The analysis compares the electrified delivery vehicle fleet with the conventional one in terms of the total energy cost and the CO2emissions for different scenarios of grid energy production sources and local RES penetration levels. Chapter 7“Conclusion” outlines the main results and the following major contributions of the doctoral thesis: 1) HEV/EREV control strategy based on combining a rule-based controller with an equivalent consumption minimisation strategy; 2) EV fleet aggregate model that faithfully describes the dynamics of EV connection to the grid, and which is parameterised based on the main features of collected driving cycles; 3) dynamic programming-based charging optimisation algorithm based on the aggregate EV fleet model, and a heuristic algorithm of distributing the optimised aggregate charging power over individual EVs.