## PhD Thesis

14/07/2022

Non interconnected island power systems (NIIPS) are characterized by high operating costs and restrictions in the renewable energy sources (RES) penetration, despite the high potential that might exist in them. Many island system operators have installed or consider pioneering projects that comprise novel control techniques and energy storage to increase the renewable energy penetration and decrease the operating costs.The present Ph.D thesis focusses on the operational security of non interconnected island systems under high RES penetration levels. Initially, the challenges in the operation of ΝIIPS operating in high RES penetration levels are presented. In addition, the contribution of the inverter based generation in frequency and voltage of NIIPS with high RES penetration is highlighted. In this thesis, a grid forming scheme for the central battery energy storage unit is proposed which can provide voltage control, emulated inertia, fast frequency response service. Its main advantage, compared to a grid following inverter, is that it ensures smooth transition to 100% renewable energy sources penetration. Secondary frequency and voltage control techniques are proposed, as well as an innovative control scheme for automatic generation control based on model predictive control. The different control schemes are evaluated in a Control Hardware in the loop testbed.The new ancillary services by the inverter based generation and the altering operating conditions in the island system require new techniques for the assessment of dynamic security. A novel data driven technique, which is based on the optimal classification trees, is proposed for the extraction of linear rules that predict the system security with high accuracy. This technique is compared with classifiers from other supervised machine learning techniques.The evaluation of data driven classifiers of dynamic security only on a scenarios set might not provide a safe assessment on whether the classifier rules approach accurately all the actual security boundaries. In this PHD dissertation, several cases of data driven classifiers are presented that have with high accuracy on the test set but can not achieve similar levels of accuracy when applied for the calculation of control commands. At the same time, an island system operator would not easily trust a black box classifier with no link to the actual physical representation of the system. Main contribution of this work is the design and application of a novel method that for the first time links a classifier’s performance with a dynamic representation of the frequency dynamics, in the same mathematical formulation, in order to calculate adversarial examples or prove that the data driven classifier rules express a subset of the secure space that is defined by that specific frequency dynamics representations. Through this methodology a secure assessment of the operating conditions according to the dynamic frequency security is guaranteed. Finally, an iterative training procedure is proposed, that leads to classifier rules that guarantee the frequency security, which can be applied in the economic operation of the island.The island’s dynamic security and the impact of the new ancillary services by the inverter based generation should be also taken into account in the economic operation of the system. An innovative formulation of the unit commitment problem is proposed that uses the optimal classification trees rules to force constraints on the frequency dynamic security. To deal with the uncertainty in the system operation, introduce by the RES in their power production and provision of ancillary services, as well as from the system load, a robust optimization formulation is proposed for the day ahead unit commitment problem.

**Utlilization of smart meter data for distribution grid operation**

12/07/2020

Modern power systems, and especially distribution networks, are now facing a set of promiscuous challenges, originating from technological developments and changes in human needs. One of these technological developments is related to the advent of smart meters, which are probably the most important interfaces between humans and the power system. The data acquired by these devices encapsulate a large portion of human behavior itself.Electricity theft is also a result of specific human behaviors taking place inside different social and economic spheres. This doctoral thesis aims in the development of novel concepts and algorithms for detecting electricity theft using smart meter data in order to reduce non-technical losses in distribution networks. A detailed review of non-technical loss (NTL) detection methods is first performed. A new taxonomy is proposed, where electricity theft detection methods are categorized according to the type of algorithm utilized. Additionally, analysis of parameters such as data requirements, extracted features, performance metrics, response time etc. is performed in order to better understand the different issues regarding NTL detection systems. Most NTL detection systems make use of data analytics and machine learning methods as their core operation algorithm. Therefore, part of this thesis is dedicated in the analysis of such methods and the problems rising during their design and implementation. Attention is given to extracting features with strong discriminative capabilities, the use of which will enhance the relative classifiers’ performance. One of these classifiers is the Support Vector Machine (SVM). This thesis provides a detailed description of the utilization of SVM for detecting NTL, but also explores the possibility of using unsupervised methods. Unsupervised methods do not require labeled samples, and thus their application can be easier, compared to supervised methods such as the SVM. An important disadvantage of data oriented methods is that they don’t take into account the physical laws governing power systems. The focus is thus moved to network oriented methods, which make use of voltage and power measurements as well as power system analysis in order to detect electricity theft. The first algorithm developed bases its operation on the calculation of distribution network voltage sensitivities. Voltage sensitivities express the relationship between voltage and power which is disrupted by electricity theft behaviors. The second algorithm utilizes optimization techniques and is inspired by the solution of traditional optimal power flow problems. Furthermore, these algorithms are combined with the SVM previously described, thus forming a hybrid NTL detection system. The aforementioned NTL detection methods exhibit high performance but require smart metering data in order to achieve it. In reality, such data may not be available. This is true for Greece and other countries, who still mostly rely on electromechanical meters for measuring energy consumption. In fact, in the case of Greece, energy is measured every four months, making the application of most NTL detection methods extremely difficult. This thesis tackles exactly this problem, by designing and implementing a rule-based algorithm for detecting electricity frauds. In addition, a software tool is designed which acts as a decision support system aiding the distribution network operator on choosing which consumers to inspect. The final part of this thesis tackles the problem of topology identification in distribution grids, by proposing a new methodology to accurately detect the tree structure and line resistances/reactances of radial networks. The algorithm first estimates the network voltage sensitivity matrices from active power, reactive power and voltage measurements. Smart meters are assumed to be installed only on the leaf nodes, and not on each network bus. The calculated sensitivity matrices are fed to a clustering algorithm, inspired by phylogenetics, which extracts the final network topology. The proposed methodology is assessed over numerous random networks and also by applying it on the optimal battery energy storage placement and sizing problem when network topology is not available.

**Contribution to operation of DC systems in microgrid and metropolitan railway network applications**

20/12/2019

The use of Direct Current (DC) technologies for electricity distribution networks is of particular international interest because of the technologies used in both production and consumption level. This interest is fueled by the need of achieving high targets in terms of share of renewable energy sources in addressing total demand. It is a fact that a large number of renewable energy sources and storage units such as photovoltaics, batteries, etc. operate with a direct voltage. Therefore, integrating these devices into DC distribution network architectures through appropriate DC/DC converters is an attractive option in terms of increasing overall performance due to the reduction of intermediate conversion steps. These systems generally provide improved reliability and easier control compared to AC alternatives, given the lack of constraints such as reactive power balance, synchronization problems, and so on. In this thesis two applications of Low Voltage DC (LVDC) systems were studied. The first category is about microgrid topologies and issues related to their control and operation while the second category is about electric railway networks operating in DC (Metro, Tram) as well as the benefits resulting from management of the generative energy producing during the braking phase of the trains. Firstly, the history of the DC networks and the basic applications of LVDC systems are presented. An introduction to the concept of microgrids and electric railway networks is performed and their basic characteristics are presented. Reference is made to the basic control architecture (hierarchical control) concerning the operation of microgrids as well as the basic islanding detection methods that appear in literature for the operation of DC networks. In this thesis a novel islanding detection method is developed based on the operation of a controllable load in parallel to the central switch of a DC microgrid. Additionally, the issue of smooth transition of a DC microgrid from stand-alone to interconnected operation is examined. Modeling of the individual elements of the studied microgrid is presented and simulations are performed based on variations of the IEEE 1547 and UL1741 standards. The performance of the proposed control strategy is evaluated through simulations, Control Hardware-in-the-loop tests and fully hardware experiments performed at the Electric Energy Systems Laboratory of National Technical University of Athens. The second section of this thesis is about electric railway networks operating in DC (Metro, Tram) as well as the benefits resulting from management of the generative energy producing during the braking phase of the trains. In railways, trains operate either as loads (acceleration - constant speed phase) or as generators (regenerative braking of trains). The energy produced is consumed by other trains running on the network or consumed locally on on-board dump loads (resistors). The produced energy cannot be returned to the distribution network since in most cases the traction substations consist of uncontrollable rectifiers (diode devices) that do not allow bi-directional power flow. Managing the energy produced leads both to reduction of losses and to voltage regulation in the railway network. In the context of this thesis, management of the generative energy is studied using: 1) Bidirectional traction substations (controllable AC/DC inverters), 2) Storage units (installed at fixed points in the railway network). Additionally, the influence of train schedules on the electrical characteristics of the railway network (substation voltage, line currents, etc.) are studied. The performed analysis include modeling of individual network elements such as electric traction substations, electric trains as well as network modeling to calculate the voltage developed between the return line and the earth (rail to ground voltage). Extensive reference to storage technologies encountered in railway applications, is made. The thesis examines both local controls and coordinated control of the units (bidirectional substations and stationary storage devices) in order to optimize network voltage regulation of the trains. The network studied is a modified version of the Metro railway network of Thessaloniki. Finally, the study of these issues is carried out using MATLAB simulation tool, in which a simulation program of the electric power flow is created in DC power networks.

**Contributions in the development of wireless inductive charging systems for electric vehicles**

30/05/2019

Inductive charging, which allows the wireless transfer of energy between the Electric Vehicle (EV) and the charging station, appears to be a very promising charging solution. Apart from the ability to transfer energy in a contactless way while the EV is parked over the station (i.e. static inductive charging), energy can also be wirelessly transferred while the EV moves on the road (i.e. dynamic inductive charging). The present PhD thesis aims to study inductive charging systems for EVs, while also proposing ways to increase the transferred power and the system’s efficiency. In this scope, a review of the compensation methods, the magnetic couplers and the control schemes is initially presented, considering the case of static inductive charging. Moreover, the basic operational principles in the case of dynamic inductive charging are studied, while examining the necessary modifications in the operation of the system when it is expected to operate with the EV moving over the station. Additionally, in the case of dynamic inductive charging, the bifurcation phenomenon (indicating the maximization of the system’s efficiency in frequencies different than the one maximizing the power transferred to the EV battery) is studied not only according to the operational frequency but also according to the movement of the EV. Moreover, the selection of the compensation capacitors’ values is examined according to the variations of the self-inductance values of the magnetic coupler’s coils (due to the movement of the vehicle over the station), in order to achieve a high transferred power at an increased system efficiency. In order to achieve a high power transfer and efficiency in dynamic inductive charging not only do the compensation capacitors need to be appropriately selected, but the optimal values of all the system’s parameters shall also be optimally defined. In this respect, an optimisation method is proposed aiming to define the optimal values for the operational frequency, the values of the compensation capacitors as well as the load resistance in a dynamic inductive charging system. Taking into account the requirements for the implementation of this optimisation method, a control scheme is also proposed that does not require any additional detection mechanisms to identify the position of the EV over the station, while eliminating communication requirements between the primary and the secondary side of the system. The effective operation of the proposed control scheme is validated by an experimental setup, comprising a circular magnetic coupler, the required compensation capacitors and the inverter implementing the suggested control scheme. Considering the electric grid side, inductive charging technologies can allow the wireless transfer of significant amounts of power to the vehicle’s battery. In this respect, a methodology is introduced for grid impact analysis of both static and dynamic inductive charging technologies in distribution networks, while also defining the maximum deployment level of inductive charging stations that does not violate technical grid constraints. Additionally, in case of a vast deployment of inductive chargers, modifications in the control system of the charging stations are examined in order to develop the necessary energy management system that will effectively eliminate issues which may arise in the grid operation.

**Development of distributed algorithms for the control and operation of smart distribution grids**

12/03/2019

Modern power networks face major challenges and newly posed problems for which alternative solutions are examined, driven by smart grid technologies. However, the rapid increase of networked intelligent electronic devices and data volume in Active Distribution Networks lead the way for the investigation of innovative solutions that would achieve efficient and robust control by dispersing the solution of the centralized problem at the level of each elementary control unit, ensuring at the same time the data privacy of the users.

This thesis aims at developing and investigating distributed techniques for the operation and control of Smart Distribution Grids. These methods require the exchange of information only between adjacent nodes (peers) and achieve the optimal solution without the need for central coordination. Initially, an overview of the most common distributed optimization models is given, together with a detailed literature review of the studies related to the proposed methods.

First, the problem of Economic Dispatch for Distributed Generation (DG) units is tackled, for which the Local Replicator Equation model is employed. The basic algorithm integrates the technical limits of the units and the generation-demand equality constraint and is thereafter extended to optimality take into account the active power losses. It achieves this by calculating in a distributed way the Loss Penalty Factors of the DG units.

Afterwards, the problem of load flow is examined. A distributed methodology is developed which is based on the Newton-Raphson algorithm in combination with a distributed protocol for calculating the correction values of voltages and angles. The method exploits an important property of the Jacobian matrix of the power flow equations.

The same property is used for reaching a distributed solution for the problem of voltage control which is formulated as a distributed optimization problem that aims at minimizing the active and reactive injections of the grid’s controllable resources, that will alleviate any voltage violations. The method uses a gradient descent algorithm and an optimization method that performs the inversion of the Jacobian matrix of the power flow equations in a distributed manner.

The accuracy and the effectiveness of each developed method, is investigated and evaluated through a series of simulation scenarios, first in simplified distribution grids with a few nodes and then in larger networks.

In the final chapter, the applicability of blockchain technologies to smart grid problems is investigated, by identifying their potential benefits and limitations. Specific categories of energy related problems are considered particularly relevant for handling them with blockchain technologies. In this direction, two scenarios of decentralized applications are designed and implemented, the first one pertaining to the automation of the electricity supply and the second one to the facilitation of decentralized energy markets in the context of energy communities.

**Contribution to assessing and enhancing power system resilience**

20/11/2012 – 10/07/2018

Natural disasters can cause significant damage to power systems. These disasters are geological and meteorological phenomena, such as extreme weather events, earthquakes, floods and wildfires. In particular, such weather events in the past few years have resulted in major power disruptions around the world, leading to unplanned electricity interruptions lasting from a few hours to a few weeks. Despite the low probability of such events, their high impact on power systems and their financial cost make the development of models necessary in order to predict them, to assess their impact and finally to deal with them.The present Ph.D dissertation focuses on power systems resilience enhancement against extreme weather events and wildfires. Resilience is the grid’s ability to withstand extraordinary and high-impact low-probability events that may have never been experienced before, rapidly recover from such disruptive events, and adapt its operation and structure to prevent or mitigate the impact of similar events in the future.Specifically, resilience indicators and metrics are proposed in order to assess the impact of extreme events on power system, quantify its resilience and evaluate various smart operational and hardening measures. Following, an online risk analysis, capable of providing an indication of the evolving risk of power systems regions subject to extreme weather events, is proposed. Furthermore, a unified resilience evaluation and operational enhancement approach, that includes the proposed online risk analysis for assessing the impact of severe weather on power systems and a novel risk-based defensive islanding algorithm, is presented.Continuing, a tri-level problem is formulated to study the contribution of the unit commitment against an upcoming extreme weather event.Finally, an optimal distribution system operation for enhancing resilience against an approaching wildfire is presented.

**Contribution to the provision of ancillary services by virtual power plants**

23/11/2007 – 10/07/2018

The widespread deployment of Distributed Generation (DG) in modern Electric Power Systems (EPS) has led to doubts regarding the future of further penetration of the former in the latter. Economical and technical reasons have already been posing limitations, even to the operation of the existing installations. Moreover, due to the complete deregulation of the electricity market and the shrinking or discontinuance of incentives supporting the development of Renewable Energy Sources (RES), the owners and/or operators of DGs have to additionally consider special issues and more complex requirements. Based on the above, an entity of both technical and economical natures has to be considered: the Virtual Power Plant (VPP). A VPP can incorporate a vast number of DG units and loads, which do not necessarily share physical connection among them, in order to fully exploit their individual characteristics and, hence, to serve market, technical and regulatory (as for the EPS) requirements and concerns. In this thesis, a set of methodologies and architectures for the provision and realization of ancillary services from a VPP has been developed, in scope of the existing grid codes and standards. In detail, the following are discussed: firm power capacity provision by a VPP, frequency control contribution by stochastic DGs, over-frequency mitigation through VPP de-loading and voltage control support of a distribution line by a VPP. It is supposed that the owner/operator of the considered VPP can neither affect the EPS nor rely on access to full information regarding the EPS status and measurements. Lastly, in the Appendix of this thesis, a methodology is proposed for managing big data and measurements logged by the actors of the VPP, since the former are required in order to serve and assess the execution/realization of the ancillary services. According to the aforementioned points, the VPP fulfills its contribution to ancillary services in a manner adequately abstract, without prerequisites or updates in codes and standards, thus can be readily integrated in modern EPS.

**Decentralized management of dispersed energy resources in distribution electricity grids for the provision of ancillary services**

22/11/2011 – 19/09/2017

Electrical grids have been primarily designed to transfer electricity mainly produced in large generation sites, and deliver it to consumers. Distribution grids constitute the subset of electrical grids, which is closer to consumption. During the last decade the operation of distribution grids has been facing a major transformation, so as to accommodate high RES penetration both at medium and low voltage grids. The operational framework of distribution grids is continuously updated and transformed exploiting the potential of Information and Telecommunication Technologies (ICT) and the newly introduced flexibility of energy resources. In the context of the thesis, nodes of distribution grids have been considered as peers, which process data locally and communicate in order to provide ancillary services.Specifically, the thesis focuses on the following topics: i) the theory of peers and gossip algorithms have been introduced into electricity distribution grids, and the acceleration of gossip algorithms’ convergence has been approached via the mathematical graph analysis, ii) gossip algorithms have been proposed for decentralized calculations of power flows in distribution networks, iii) gossip algorithms have been designed and developed for distributed congestion management, iv) a multi agent system, which integrates the proposed algorithms, has been designed and developed, v) the exploitation of smart meters and other ICT devices is proposed in order to create overlay networks towards the exchange of information and the application of distributed algorithms in distribution network management. The expansion of smart meters’ protocols has been examined and proposed.The thesis begins with a bibliographic review of the current operational and management framework of distribution grids and a detailed review of distributed optimization algorithms, followed by a description of gossip algorithm theory. Based on graph spectral analysis optimal weights are selected to accelerate the convergence of gossip algorithms. Generic gossip algorithms have been extended to calculate power flow equations for radial distribution grids. The proposed gossip algorithms are the core of a distributed framework for congestion management. Nodes are sensing the constraints’ violation at any part of the network by exchanging information within their neighborhood and they are self-triggered to resolve the violation by participation in a totally distributed optimization framework. Applications of gossip algorithms are also applied at the various voltage levels for coordinated congestion management and for local balancing of production-consumption. Moreover, a parametric estimation of telecommunication and calculation load of gossip algorithms is presented.In the second part of the thesis, a multi-agent system is designed and applied for ancillary services provision by distributed energy resources based on the proposed gossip algorithms. Finally, techniques for information coding applied in electronic meters are investigated and an expansion of DLMS/COSEM classes is proposed, in order to enable the application of gossip algorithms for decentralized management of distribution grids through electronic meters.

**Real-time simulation of the impact of distributed generation on distribution networks using digital models and hardware components**

19/10/2010 – 11/07/2017

The subject of this doctoral thesis is the use of real-time simulation and specifically Power Hardware in the Loop (PHIL) simulation, to study the impact of distributed generation on the distribution network. PHIL simulation allows the connection of hardware power equipment (e.g. photovoltaic inverter) to a simulated network. The potential future role of PHIL simulation for laboratory testing and studying of phenomena in power systems is explained. The stability and accuracy issues of PHIL simulation, the main methods for evaluating and achieving stability and methods for accuracy estimation are presented. A new method to achieve stability is proposed, which maintains sufficient accuracy, while it allows the connection of a scaled-down device to analyse the behaviour of the full-scale device. Stability considerations are provided and the accuracy improvement by using the proposed method is quantified. Next, modern functions for the provision of ancillary services by distributed generation are presented concerning voltage control (cosφ(P), Q(V) control) and frequency control (P(f) control, virtual inertia). The stability issue of Q(V) control is explained. A review of recent international standards/guidelines is performed, concerning steady-state voltage support, dynamic network support and frequency control, which includes a comparative analysis and identification of gaps. Advanced testing procedures using conventional approaches are proposed, along with PHIL test procedures. Subsequently, the development of the PHIL environment at the Electric Energy Systems laboratory of NTUA is described, which includes the protection schemes applied and preliminary PHIL experiments are presented. Α benchmark system for PHIL testing is proposed, which includes reference test procedure, setup and network. Problematic interactions between local voltage controller of distributed generation and On-Load Tap Changer (OLTC) are examined with pure digital simulations. The proposed method to achieve stability and accuracy is applied and PHIL tests are executed using a scaled-down photovoltaic inverter to analyse the behaviour of the full-scale inverter. The PHIL tests show interactions that were not visible at the pure digital simulations, which demonstrate the value of the PHIL approach. The model of the voltage controller of the inverter is validated by comparing the results of the PHIL tests and the pure digital simulations. PHIL tests on frequency control are performed to study the impact of droop control and virtual inertia. Next, PHIL simulation is employed for laboratory education of students on important topics of the operation of the modern power system. A double PHIL setup is developed, creating two work benches, which provides hands-on experience to the students in small groups and allows experiential learning. The assessment of the laboratory exercises by the students was clearly positive. Finally, the electromagnetic interference from inverters of distributed generation on electronic meters is analysed, which can result in energy measurement errors. A detailed review of standards/guidelines is performed, which highlights the gap in the 2 kHz – 150 kHz range and reports recent developments and remaining issues. Laboratory setups for testing immunity and emissions are described and immunity tests on electronic meters and emissions tests on inverters are performed.

**Optimal management of electric vehicles for their efficient and effective integration into electricity networks**

06/10/2009 – 11/04/2017

The electrification of the transport sector requires the interconnection of the Electric Vehicles (EV) with the electricity grids utilizing appropriate charging infrastructures. The EV energy needs should be served as domestic consumption in a non-discriminatory way. The difference between EV and domestic demands lies in the fact that the charging demand is not static due to EV mobility and it presents intense spatial and temporal volatility. From the grid perspective, the EV charging demand is an additional load which can significantly modify the system load curve and affect the way electricity networks are operated and managed, in respect to the EV deployment level and the charging policy. The present PhD thesis aims to analyse the impact of EV deployment on the operation of electricity grids and to propose EV management algorithms that enables their efficient integration in the electricity grids in respect to the outcome of the grid impact analysis. The objective of an EV management is network dependent and it is directly linked to the technical and operational characteristics of the respective network under study. Thus, a holistic grid impact analysis (generation system, transmission/distribution networks, non-interconnected systems) is necessary in order to identify the potential operational market/network issues that EV share will provoke examining different EV penetration scenarios and EV deployment strategies. The two main goals of the proposed EV management algorithms in the present PhD thesis are: the maximization of the EV hosting capacity of distribution networks considering the current grid infrastructure as well as distributed renewable energy and the integrated market participation of the aggregated EV batteries’ capacity, taking into consideration the bidirectional power flow between the electric vehicles and the electricity grid. The dynamic EV behavior due to their mobility requires the development of advanced smart charging solutions aiming to serve EV charging needs balancing EV users’ preferences, network operational constraints and electricity market parameters. The implementation of decentralized management algorithms is highly volatile environments can be more efficient compared to centralized ones. Thus, the present PhD thesis focuses on the design and development of decentralized EV control algorithms. The performance of the proposed decentralized EV control algorithms is compared to the one of the respective centralized methods concluding that distributed algorithms provide results of similar quality requiring shorter computation time under high EV share.Last but not least, a standard custom-made charging station and an interoperable back-office system for charging stations network has been developed enabling the performance of compatibility tests to e-mobility standards of different commercial e-mobility technologies and the implementation of innovative EV management algorithms in laboratory or real environment.

**Optimal management of distributed energy resources**

23/11/2007 – 11/04/2017

The advent of new types of distributed resources such as flexible loads and local generating units poses significant challenges to the operation of the electricity system in total and to the incorporation of such resources in the market procedures. The present PhD thesis focuses on analyzing the impact of the operation of distributed resources as those mentioned above to the electricity system operation and the manner in which these resources operate. To this end, an hourly unit commitment and economic dispatch algorithm is applied in order to test the impacts of load shifting in the operation of the electricity system in terms of energy quantities and in economic terms. Furthermore, a bilevel model is formulated and put to use in order to study the interaction of an Aggregator, responsible for representing various distributed resources in the market operations, with his customers. Aiming toward minimizing the energy procurement cost, the Aggregator optimally selects the retail prices announced to his customers as well as the amount of energy acquired from the network. Based on these prices, the entities possessing any type of distributed resource optimally select the amount of energy to be produced, demanded or curtailed. The full problem of the participation of the Aggregator in the market procedures is studied next. An appropriate bilevel formulation models the decision making process of the Aggregator that is no longer influenced only by the characteristics of the distributed resources under his control. The outcome of the market clearing process, to which the Aggregator submits appropriate production bids and load declarations, affects his decision regarding the formulation of such bids and declarations. Last but not least, the problem of a regulatory authority responsible for the long-term scheduling of the electricity system is formulated as a bilevel problem, with a view to optimally select the incentives given to investors in wind energy projects that minimize the energy procurement cost.

**Contribution to Management of Microgrids through the Integration of Renewable Energy Sources and Co-generation Systems of Heat and Power**

29/11/2005 – 09/12/2014

The installation of more and more units of dispersed generation (DG) close to consumers leads to a new era for the Electrical Energy Systems. This doctoral thesis aims to develop proper methodologies, tools, software and analysis techniques for the quantification of the dispersed generation’s advantages, giving emphasis in Combined Heat and Power production (CHP) and Renewable Energy Sources (RES), whether they operate independently or in a coordinated way, thus forming a Microgrid. The transformation of the deterministic equations that govern these studies to their stochastic form as well as the development of methods to solve the probabilistic load flow in order to manage the uncertainty isnecessary.

More specifically, a methodology was developed, with the assistance of the sensitivity analysis method, for determining the sensitivity coefficients, for those quantities of the grid that are used for the quantification of the main economic benefits because of the augmented DG penetration. Those economic benefits, through an implementation in a specific grid, arise from postponing the investments in new grid components such as transformers and cables, from the reduction of the total active power loss, from avoiding buying energy based on the system’s marginal price during peak hours and from the potential reduction in electricity price. The improvement of the voltage level on the system’s buses was also examined.

Moreover, suitable calculation software for minimizing the operational cost in the Microgrid in which the energy and pricing policies are applied was also developed. The studied low voltage (LV) grid consists of consumers (loads) and energy microsources such as photovoltaics (PV), wind turbines (WT), microturbines (MT), fuel cells (FC), micro_combined heat and power units (Micro_CHP), electrical vehicles (EV). Policies that regard the augmented penetration of DG units were also applied. Furthermore, the Microgrid was sufficiently studied with variations in the pricing policies relevant to its operation, where the variable of energy demand elasticity as to the price was introduced. A number of logistic pointers, which are extracted for each one of the operational scenarios, were created, in order to manage the extracted information, so as to compare the scenarios and, consequently, the policies that can be implemented in a LV network, between them. So, by evaluating comparable scenarios, the policies that contain greater viability and realization potentiality are presented. The modeling was done in such as way as to ensure the maximum possible autonomous operation from the upstream network.

In addition, a viability study of a Micro_CHP investment for a residential complex compared to the conventional coverage of its energy needs was conducted. Then, a probabilistic model of a Micro_CHP unit in association with the external ambient temperature and the characteristics of the area to be thermally covered, so as to find the distributions of both the electrical output and the thermal power using the HPR ratio were examined. The proposed probabilistic model was used to find the hourly (or the aliquot of the hour) average thermal and electrical demand of the buildings. In larger electrical systems, the procedure to find the optimal load flow with environmental constraints and with CHP extraction units in Heat Match operation was successfully developed, highlighting these units as main players.

Then, finding the solution for the stochastic multi objective optimization problem in a Microgrid is achieved by taking into consideration that the functions of the expected operational cost and the risk functions (expected deviations of the electric and thermal power production) that will be minimized using the weighted sum are conflicting. Expanding and solving the known stochastic model of economical distribution using CHP extraction units so as to include wind parks, air pollutants and electrical system’s safety function due to augmented wind penetration is sufficiently dealt. This is how the operator has a strong decision making program so as to achieve more accurate solutions and estimations for a power system with RES penetration, under environmental constraints.

All in all, flexible computational software was developed for most of the arithmetic and analytical solution methods of the probabilistic load flow in electrical systems with RES penetration in order to make possible the comparison of the results and their efficiencies having Monte Carlo method as comparison base. In particular, the implementation of linearization of the AC load flow equations combined with the Gram-Charlier expansion or Cornish-Fisher method in the grid of Crete extracted reliable results in a very short time frame compared with the other examined methods. Moreover, the study of probabilistic load flow in a LV network with a Micro_CHP unit whose output (electrical and thermal power) is derived from the probabilistic model is fully conducted. The study of the effect of the external ambient temperature in voltage and power flow in the grid’s lines can be useful in cases when there many Micro_CHP units that operate similarly have penetrated.

**Contribution to transmission expansion planning and transmission fixed cost allocation in electric power systems**

23/11/2007 – 17/07/2013

The deregulation of electricity markets has resulted in the separation of electricity transmission systems from the former vertical integrated utilities. In the deregulated environment, pricing of this natural monopoly should promote the economic and secure power system and market operation, allocate in a fair way the regulated revenue of the transmission owners to network users and provide the right economic signals for new transmission, generation and energy efficiency investments. The main objective of transmission system expansion planning in deregulated electricity markets is to provide competitive and non-discriminatory market conditions to all market participants while maintaining power system reliability and security to acceptable levels. The subject of this thesis is: i) the implementation and evaluation of various transmission fixed cost allocation methods to network users and ii) the development of new methods and optimization tools to address the uncertainties involved in power system transmission expansion planning problem in deregulated electricity markets. In this thesis, the implementation of various approximate power flow tracing techniques and transmission fixed cost allocation methods is initially examined in a pool-based electricity market. Subsequently, a new method for allocating the annual fixed cost of each transmission facility is developed based on an optimal (smallest possible in terms of N-1 security) transmission facility capacity approximate use. The use of each facility is calculated using the generalized distribution factors for that system snapshot that requires the maximum optimal transmission facility capacity over an annual system operation. The proposed method can identify and charge accordingly the main beneficiaries of the existence of new and existing transmission facilities in various power system operating conditions. At the end of the thesis, the proposed transmission fixed cost allocation method is compared and implemented in parallel with marginal pricing of transmission considering either the whole network’s annual cost or only the fixed cost of the new installed facilities.In the second part of the thesis, an improved harmony search algorithm is initially implemented and evaluated in the mixed integer static transmission expansion problem. A new formulation of the probabilistic transmission expansion problem is then proposed by introducing an upper limit to possible load shedding during system peak. This probabilistic problem is solved by using Benders decomposition technique along with Monte Carlo simulation for modeling input data uncertainty (peak load forecast, transmission facilities availability and wind farms production). Finally, two different cost-benefit formulations for the objective function of the transmission expansion planning problem in deregulated electricity markets are proposed depending on whether the final solution of the transmission expansion problem will be implemented as merchant investment or as an investment that aims to improve the economic operation and reliability of the system. These two problem formulations are addressed by applying the proposed improved harmony search algorithm using marginal pricing of transmission that incorporates the cost of N-1 security in system locational marginal prices (LMPs) and considering the optimal capacity of network facilities as a criterion for selecting the candidate branches for expansion.

**Wind power forecasting using neural networks and fuzzy logic techniques**

18/11/2003 – 23/02/2012

In the present doctoral thesis, two advanced wind power forecasting methodologies are developed. The methodologies are based on techniques of artificial intelligence.

The two wind power prediction models that were developed, give precise forecasts of a wind farm production for the next two or three days, based on the numerical weather predictions that are provided by meteorological models with high resolution. Beforehand, the processes that they follow the meteorological models are analyzed

and the factors that influence the relations of numerical weather predictions with the wind farm production. These relations compose the problem of wind power forecasting.

The structure of the first wind power forecasting model that will be presented in the present thesis is based on the quality of the numerical weather predictions. Firstly, the

weather forecasts’ reliability are estimated with a fuzzy inference module and depended on the numerical weather predictions quality, one of the radial base neural

networks of the model are activated. The structure of the second proposed wind power forecasting model is based on the atmospheric conditions at the prediction time. The atmospheric conditions are first classified via a self-organized map. Then radial base function networks provide a first estimation of the wind power production. In the

following step the uncertainty of the numerical weather predictions is estimated. At the end, radial base neural networks correct the initial wind power estimation base on

the uncertainty of the numerical weather predictions. At the second wind power forecasting model a methodology was integrated that modifies the wind power point

predictions to predicted wind power densities.

The validation of the proposed wind power forecasting models is done in off-line and in on-line mode in real wind farms with different characteristics. Moreover, the online performance of the models is compared with the respective performance of two different state-of-art wind power forecasting models in five wind farms on the same

evaluation period.

**Contribution to short-term and mid-term load and energy forecasting based on fuzzy logic**

18/11/2003 – 23/01/2012

The present thesis is focused on the investigation and the study of the problem of load and energy demand forecasting of power systems in short-term and midterm system using fuzzy logic. Firstly, a brief introduction to load and energy forecasting is made, the significance of them is distinguished and their applications, as well as the factors affecting the demand for electricity are analyzed. In addition, the various loads and energy forecasting models referred to bibliography are analyzed. In order to satisfy the demand of short-term load forecasting, a new algorithm is developed for forecasting daily load curve for the interconnected power system using fuzzy logic for the next and the one after that day while it has been performed the calculation of the provision fine with respect to corresponding days in accordance with relevant regulations. This algorithm is optimized as far as the choice of it, the characteristic parameters of the participation function as well as the various models of inference. This model was also studied in terms of behaviour towards the years of education. It is implemented for the year 2008. In order to determine the confidence interval, the basic methods for standard deviation calculations for the artificial neural networks have been recorded. After that the sampling method has been modified for case of the fuzzy logic algorithms. An innovative method for analytical calculation of the standard deviation has also been developed, which combined with the probability density function of the error leads to the calculation of the confidence interval. The respective results are compared with different criteria such as the empiric coverage, the quality interval and the relative confidence interval, where the superiority of the innovative method is proved. The proposed fuzzy logic algorithm was suitably amended in order to be applied to autonomous power systems, such as to the power system of St. Efstratiou Island. The algorithm calculates the daily curve of the next day having as data a minimum number of 100 days (due to lack of data). Therefore, based on the principles of fuzzy logic a corresponding model of midterm energy forecasting was created. This model was applied to predict the energy required for the next three years for the whole Greek system and for individual categories of consumers. The results were clearly better than those obtained by conventional methods of energy forecasting and a little better than corresponding forecasting of artificial neural network models. Finally, the general conclusions arising from the creation and implementation of the short-term forecasting models and midterm energy forecasting are recorded, while prospects for further research in relation to the above fields are mentioned. At the end of the thesis there is a detailed description of the basic elements of fuzzy logic, fuzzy sets, basic operations among them and a brief introduction of artificial neural networks and back-propagation algorithm is also presented.

**Wind turbine modeling for dynamic studies during high wind power penetration**

17/04/2007 – 13/12/2011

Ιncreasing wind power penetration in modern power systems is posing serious challenges to wind turbine technology regarding wind turbines’ capability to operate according to the restrictions specified by system operators. Numerous regulatory and control functions traditionally offered exclusively by conventional units are nowadays required by wind turbines connected to the grid. The scope of this PhD dissertation is to develop dedicated control and design techniques for wind turbines enabling them to behave as active components of power systems. During the dissertation emphasis was given to two basic types of variable speed wind turbines – the doubly-fed induction generator wind turbine and the direct driven wind turbine with synchronous generator (electrical excitation or permanent magnet) and full converter. The models are used to investigate the interaction of wind turbines-wind farms with the grid during normal operation and during transient events, such as faults in the grid. In the first step, dynamic models are developed representing the aerodynamic and mechanical behavior of the wind turbine as well as the complete electrical system – generator, converters and connection to the grid. The models were developed in simulation softwares such as Matlab / Simulink and DIgSILENT Power Factory. Afterwards control strategies are designed for each type of wind turbine for normal operation in order to investigate the effect of wind power fluctuations in the electricity system due to fluctuations in wind speed and additional control techniques for the operation during transient phenomena in the grid. The first type of transient phenomena refers to the fault-ride through capability during voltage dips for the two types of wind turbines and the possible contribution to the voltage control. The second type refers to events with significant impact on the active power balance and system frequency. Frequency controllers are designed for wind turbines to be able to contribute to the primary frequency control in a similar way to conventional units. Control strategies do not only refer to individual wind turbines but also to wind farms with a rated power that approaches that of conventional power plants. The controllers at the wind farm level regulate active power production of individual wind turbines based on signals received from the system operator for various functions, such as active power reserve, inertial response, droop control and limits of rate of change of active power. The models developed are then used to investigate the operation of wind farms in modern power systems, with emphasis given on non-interconnected systems, such as island systems. The characteristics of isolated systems pose special requirements for grid support services on behalf of wind farms. The methodology approach includes investigation of the operation during wind power fluctuations due to the stochastic nature of wind and analysis of the interaction between wind turbines and the grid during transient phenomena that affect either the voltage or the frequency of the system. The steps of investigation establish a methodology to estimate the limits of wind power penetration as well as other parameters, such as the amount of active power reserve required by wind farms during primary frequency control. Although the quantitative results produced with the specific system under study (Rhodes power system) can not be generalized, it is clear that the control strategies developed and presented in this dissertation provide with the possibility of extending the limits of wind power penetration without violating the dynamic security margins of the system.

**Machine learning contribution to dynamic security control of power systems**

18/11/2003 – 08/06/2010

Security is defined as the capability of maintaining the continuous operation of a power system under normal operation and following significant perturbations. As the increase in electric power demand outpaces the installation of new transmission and generation facilities, power systems are forced to operate with narrower security margins; therefore fast and reliable assessment of the system security is necessary. In the field of Dynamic Security Assessment (DSA) much attention has been paid to preventive, as well as corrective control. Preventive control refers to a set of actions that are applied when a potentially dangerous violation is detected through DSA. Corrective actions are applied to offset a security violation after the occurrence of a threatening contingency.

The scope of this dissertation is to apply machine learning methods to the dynamic security assessment and control of power systems. Machine learning methods are seen to have features that can bring benefits in comparison to analytical methods. Once developed, they provide solutions very fast. Furthermore they have the ability to recognize, if a system condition has previously occurred and predict its security accordingly. Similarly, if properly designed, they can adapt to new conditions by learning from situations previously seen. Finally they provide a high degree of discovery, i.e. they have the ability to uncover salient, but previously unknown, characteristics of, or relationships in, a system

Decision Trees (DT) are applied to propose security rules, which can be used for corrective load shedding strategy as well as security constrained economic dispatch determination. A modified algorithm for the construction of Decision Trees, that utilizes genetic algorithms to formulate characteristic groups of load buses, is also proposed and applied to derive load-shedding schemes. The DT method is further developed and an automatic learning hybrid method for corrective dynamic security is proposed, based on a Self Organized Map (SOM), that classifies the power system’s security state according to its load profile. DTs nested in nodes with mixed security states are applied to investigate further their security status. Radial Basis Function Neural Networks (RBFNN) and Support Vector Machines (SVM) are applied on DSA, tested with various feature selection techniques. A hybrid method of RBFNN and Particle Swarm Optimization (PSO) is developed for the definition of remedial actions as well as appropriate preventive control. A Reinforcenment Learning framework is proposed for the determination of a spinning reserve policy of an autonomous power system. The algorithm is trained online and compared to deterministic spinning reserve policies. Finally a machine learning method that combines PSO with a modified SOM algorithm is developed for the determination of control islanding of power systems

The methods are applied on a realistic models of Cyprous and Hellenic power systems, the MicroGrid and MultiMicroGrid as well as IEEE test systems.

**Contribution to the control and simulation of low voltage power systems with distributed generation**

26/11/2001 – 02/06/2009

The drivers of dispersed generation urge its extension to the low voltage network, for which it is contemplated installation of numerous small capacity generators with diverse technical characteristics that are connected using voltage source inverter at their output. Sources and loads as a whole will constitute a single entity for the upstream network, also called microgrid, which shall be able to operate autonomously as a self regulated island in the event of problems to the upstream network, a fact that entails higher consumer reliability and extension of the dispersed generation production. Islanded operation depends on paralleling the voltage source inverters of the distributed sources that are capable of regulated power output. Inverter paralleling using only local measurements is implemented with proportional control of frequency and voltage with active and reactive power output respectively. The installation of storage devices is necessary due to the fact that the created AC system will lack any form of inertia.

This work focuses on the autonomous operation. The local control of the sources through their inverter is investigated so that stable operation is secured. Control methods are proposed so that typical system functions such as load or local generation fluctuations take place smoothly without power oscillations, which is translated to voltage and frequency regulation. The inverter at the output of the sources has implications, except from the operation, to the modeling itself that is needed for the dynamic analysis of the system. Therefore the transient stability methods, as applied to a system with synchronous machines, are adapted for application to an isolated system with voltage source inverters, accounting also for the asymmetries of low voltage network. So, the simulation of the dynamic behavior of the system can be carried out, testing various control strategies according to the characteristics of the power output change of each source.

The behavior of the system depends on the magnitude and angle of the impedances that interconnect the voltage source inverters as well as on the control parameters, which are the droops of the frequency and voltage of the inverters with powers and the delay involved in this change. The factors that define the scope of the controlled system parameters are examined first. Controlling frequency with active power has the advantage that the active power production of each source is defined exclusively with the control parameters, irrespective to the position in the network. The influence of the resistance to reactance ratio is considered and specific means to improve the control performance is proposed. System stability is thoroughly analyzed considering two inverters in parallel and applying control methods of the classical control theory. The network dynamic is also included in the analysis on account of the fast response of the inverter. A specific compensation is proposed to improve the relative stability so that the defined response specifications are met for the anticipated system parameters.

A large percentage of the load in the low voltage is non linear, thus if the control is confined to the fundamental frequency, voltage distortion may prove prohibitive for the islanded operation. The proposed method employs closed loop control of the terminal voltage and feed forward of the load current for each voltage source inverter, thereby achieving good voltage quality and harmonic sharing among the inverters in proportion to their capacity. The control implementation is in the time domain using instantaneous values thus with overall harmonic compensation. The application of the control method is tested with simulations.

**Contribution in operation scheduling of distrubition networks with high penetration of distributed and renewable energy sources and stroage devices**

17/10/2002 – 15/07/2008

The vision of EU for future grids is that they must ensure secure and sustainable electricity supplies, take advantage of new technologies and comply with new policy imperatives and changing business frameworks. This thesis deals with scheduling of distribution networks with increased penetration of distributed, especially Renewable Generation and Storage. This topic constitute an interesting research area for achieving the goals of the EU vision. In this thesis, a methodology for assessing the impact of Distributed Generation (DG) in the emissions avoidance of the upstream network has been developed. Special focus is given on the fact that DG production displaces the last dispatched units of the upstream network. Based on this methodology, a software for evaluating Combined Heat and Power plant impact on the emissions avoided has been developed.

Interconnection of small, modular generation, storage devices and controllable loads to distribution networks forms a new type of power system, the Microgrid. A Microgrid can operate either interconnected or isolated from the main distribution grid as a controlled entity. In this thesis, the operation of the Microgrids Central Controller (MGCC) is presented. The controller aims to optimize the operation of the Microgrid during interconnected operation, i.e. maximize its value by optimizing

production of the local DGs and power exchanges with the main distribution grid. Two market policies are assumed including Demand Side Bidding (DSB) options for controllable loads. Moreover, restrictions such as maintaining adequacy of the Microgrid in case of upstream fault- either taking into account low priority loads or not- are incorporated in the developed software for the simulation of the MGGC operation. Results from a typical LV study case network operating under various market policies, assuming realistic spot market prices and DG bids, accounting also for both DSB and adequacy constraints, show the potential of the suggested operation. A further step is the participation of a Microgrid in the emissions market as a means of increasing its income. The emissions assessment with the above mentioned methodology shows significant CO2 emissions avoidance and increased income for the Microgrid.

Island power systems, a special operational mode of Microgrids, face increased needs of spinning reserve especially when operating with high RES penetration. A methodology based on probabilistic techniques is proposed for evaluating the uncertainty of the load to be distributed to the thermal units of such a system, according to various levels of information regarding the errors of load and RES forecasting, as obtained by their evaluation. The proposed formulation is developed for various levels of

detail on the performance of wind power forecasting (wpf). Comparing the various levels of detail of the same wpf tool has shown significant difference in both operating cost and potential curtailment of wind power.

Intermittent power generation and high volatility of market prices are some of the many reasons to start considering connection of energy storage systems in the distribution networks. In order to evaluate the cycling effect of the storage devices, a software has been developed to simulate the operation of a storage device with the same initial and final state of charge either in open market environment or in autonomous power systems for various levels of initial content and various market prices. This software is based on an heuristic optimization algorithm and models the behaviour of lead acid batteries with modifications of the Peukert equations. Results from characteristic days of both market environment and autonomous power systems are presented. Moreover, the assumption that a storage device provides part of the spinning reserve for autonomous power systems has been assessed. Applied to Kythnos, a storage device providing partly spinning reserve and partly used for wind power management presents encouraging results for further investigation. The modeling equations for lead-acid batteries have been also utilized for the developed operational tool of remote power systems which helps in both avoiding excessive start-ups of the diesel generator and reducing the magnitude of the current to and from the storage device. Finally, the impact of various management schemes for controllable loads on the operation of autonomous power systems is discussed and topics for further research are suggested.

**Contribution of Machine Learning Methods to the Dynamic Safety Assessment of Electricity Systems**

19/12/1995 – 17/07/2006

**Contribution to the distributed control of power systems with distributed energy resources in low voltage**

26/11/2001 – 25/07/2006

**Contribution to stochastic assessment and mitigation of volage sags**

21/12/1998 – 08/03/2005

**Contribution to the analysis of thyristor controlled contribution to the analysis in transmission lines and determination of their optimal location in power systems**

21/11/1994 – 12/12/2006

**Contribution to simulation and identification of wind turbines with asynchronous generator for the development of reduced order equivalent models**

21/12/1998 – 09/12/2005

**Contribution to the transients reduction during controlled switching of high – voltage shunt reactors and shunt capacitors**

21/11/1994 – 18/05/2004

**Contribution to the optimal allocation of frequency and voltage control ancillary services in open energy market environment using machine learning methods and meta-heuristic algorithms**

21/12/1998 – 28/05/2004

**Contribution to modelling and analysis of the transient behaviour of grounding systems**

19/12/1995 – 26/11/2001

**Contribution of artificial intelligence techniques in the reduction of distribution transformer iron losses**

19/12/1995 – 13/04/2000

**Computational intelligence contribution for on line dynamic security assessment of autonomous power systems**

22/01/1998 – 18/11/2003

**Signaling Implementation and Signaling Controls for Broadband Networks**

02/04/1992 – 19/02/1997

**CONTRIBUTION OF PROBABILISTIC LOAD FLOW TO OPERATIONAL CONTROL OF POWER SYSTEMS**

1981-1986