semester- & master theses fs 2018 power systems laboratory · replacement of these assets....
TRANSCRIPT
Semester- & Master Theses FS 2018
@
Power Systems Laboratory
www.psl.ee.ethz.ch
P LPower
Systems
Laboratory
Adaptive Benders decomposition for stochasticunit commitmentSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Type Semester/Master
Background The large-scale integration of stochastic renewable generation in thepower system requires novel operational strategies and market clearing mechanismsthat account for the uncertain in-feed of those resources. An efficient method to copewith this uncertainty is stochastic optimization, which unfortunately involves a highcomputational burden when applied to real-life power systems. In order to improvecomputational tractability, stochastic optimization problems can be solved using de-composition techniques that improve solution times while guaranteeing convergenceto the global optimum.
Description This project will investigate potential improvements to state-of-the-artBenders decomposition algorithms to exploit the statistical properties of the scenarioset. Different versions of the Benders algorithm will be applied and evaluated on astochastic version of the Unit Commitment (UC) model, which is an important com-ponent in many electricity markets around the world. Tasks are to
1. Review the literature and acquire a good level of understanding of stochastic opti-mization and Benders decomposition.
2. Implement a stochastic UC model and solve it with the standard versions of Bendersdecomposition (single- and multi-cut approaches).
3. Investigate potential improvements in terms of computational time using scenarioclustering techniques.
4. Time permitting, develop a parallelization strategy to exploit High PerformanceComputing resources.
Prerequisites Interested students should have a strong background in optimizationand power systems operation; solid understanding of applied statistics is also anadvantage. Optimization models may be implemented in Python, MATLAB, or GAMSdepending on the background of the student.
P LPower
Systems
Laboratory
Asset management in distribution networksSupervisor Stavros Karagiannopoulos, [email protected]
Dr. Omid Alizadeh-Mousavi (DEPsys), [email protected]
Type Semester/Master
Background DEPsys is a Swiss energy company focused on integrating decentral-ized renewable generation (e.g., solar photovoltaic) and distributed storage into low-voltage electricity grids, and its GridEye software helps distribution system operatorsmonitor and optimize their networks. In addition to local asset monitoring systems,DSOs need a global overview of critical assets like transformers and cables. Thisproject will develop and validate models to estimate the status and importance ofsuch assets to provide improved decision support for asset managers.
Description The goal of this thesis is to investigate new approaches for distributionnetwork asset management. The developed methods will estimate the aging of theseassets and indicate when they are approaching their end-of-life. Then, risk-based assetmanagement will be developed and used to optimally schedule repair, retirement, andreplacement of these assets. Tasks are to
1. Study, develop, and implement aging models for distribution grid assets.
2. Validate the developed models using available measurement data.
3. Evaluate asset condition using the aging models.
4. Prepare a list of factors to evaluate the importance of the assets for grid operation.
5. Analyze the status and importance of assets as part of risked-based asset manage-ment.
Prerequisites The electrical engineering student should have some familiarity withelectric power systems and distribution networks; experience writing efficient andreusable code in C, Python, and MATLAB; experience working with real data; goodknowledge of power system economics; and familiarity with IEEE/IEC standards.
P LPower
Systems
Laboratory
Comparison of power system state estimationalgorithmsSupervisor Aleksandar Jovicic, [email protected]
Type Semester
Background State estimation (SE) is a vital part of modern power network moni-toring systems. It processes raw measurements from meters installed in the grid andestimates the system operating state. This estimate is normally given as a set of volt-age phasors for each bus. Based on the SE output, system operators can confidentlyexecute different network analyses and respond appropriately. Since Schweppe es-tablished the fundamentals of SE in the 1970s, researchers have worked to refine thistool and improve computational efficiency, robustness, and accuracy.
Description The student will examine and compare the performance of well-knownalgorithms for transmission system state estimation. Basic algorithms for SE will besuggested by the supervisor, but the student is encouraged to propose additional oneswhich may improve certain aspects of SE analysis. Experiments will be performed us-ing several test cases, including systems with more than hundred buses. Differenttypes of measurement units, such as remote terminal units (RTU) and phasor mea-surements units (PMU), will also be considered and investigated. Tasks are to
1. Review literature on existing SE algorithms and become familiar with the topic.
2. Write MATLAB code to generate test cases for systems of various size and withdifferent scenarios regarding measurement availability (i.e., multiple combinationsof PMUs and RTUs).
3. Implement the set of algorithms which are to be examined.
4. Compare the performance of the investigated methods in terms of computationaltime, robustness, and accuracy, and draw conclusions about their efficiency in treat-ing different types of measurements.
Prerequisites Students with a solid background in power systems and good program-ming skills in MATLAB are preferred.
P LPower
Systems
Laboratory
Convex relaxations for optimal gas flow problemSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Andreas Venzke (DTU), [email protected] O’Malley, [email protected]
Type Semester/Master
Background The integration of stochastic renewables requires flexible gas-firedgenerators to respond to renewable variability and causes increased dependency be-tween the electricity and natural gas systems. Hence, integrated optimization frame-works are needed to ensure their efficient operation. However, since gas and electricitynetworks are governed by non-convex equations, solutions obtained using non-convexsolvers may not be globally optimal. Convex relaxations can transform non-linear ACoptimal power flow (OPF) into a convex problem that can recover the global optimum.Recent literature has proposed several approaches to convexify the optimal gas flow(OGF) problem.
Description The main objective of this project is to develop a convex formulation forthe co-optimization of electricity and gas systems. Existing convexification methodsof the gas flow problem will be combined with suitable formulations of the electricitynetwork, and their performance will be compared to a non-convex benchmark model.The project will investigate the computational requirements of each convexificationapproach and assess the benefits of co-optimizing electricity and natural gas systems.The main tasks of this project are to
1. Review literature on OPF and OGF formulations and related convex relaxations.
2. Implement and assess the performance of different convexification methods of thegas flow problem.
3. Extend the formulation to include (accurately modeled) electricity networks.
4. Investigate the relaxation gap of the developed formulations and evaluate theirperformance against a non-linear, non-convex benchmark model.
5. (Optional) Formulate a multi-period optimal gas and power flow problem consid-ering temporal dependencies in gas flow.
Prerequisites Interested students should have a strong background in optimizationand power systems. Optimization models may be implemented in Python or MATLABusing MOSEK solver, depending on the background of the student.
P LPower
Systems
Laboratory
Convex relaxations for solving AC power flowproblemSupervisor Dmitry Shchetinin, [email protected]
Dr. Tomas Tinoco De Rubira (EPRI)
Type Semester/Master
Background Steady-state power flow analysis is a basis of power system planningand operation. Since the underlying physical equations that govern flows in powernetwork are nonlinear, a closed-form solution of the AC power flow problem is notpossible. Therefore, this problem is solved by numerical methods such as Newton-Raphson, which require a good starting point to converge to a solution. This meansthat if the solution algorithm does not converge, it is generally impossible to tellwhether there is indeed no solution or the chosen starting point is not good enough.Such a situation is clearly undesirable because power grids are critical infrastructureand the algorithms used for their analysis and control must be as robust as possible.One potential way of addressing this challenge is the utilization of so-called convexrelaxation techniques, which enables the application of fast and reliable solutionalgorithms to various problems arising in power system operation.
Description The main goal of this project is to develop an algorithm for solving theAC power flow problem based on convex relaxation techniques. These techniques haveonly been applied in optimization, which makes it interesting to apply them to thepower flow problem. The problem should be reformulated as a convex optimizationproblem. To do this, convex relaxations to power flow equations should be obtainedbased on methods proposed in the literature. To improve the quality of relaxations andtune the formulation of the optimization problem, the properties of the power flowproblem will be exploited. The efficiency of the proposed approach will be analyzedthrough a number of test cases representing power grids of different sizes. Tasks are
1. Literature survey on convex relaxations applied to power system optimization.
2. Formulate an optimization problem that corresponds to a convex relaxation of theAC power flow problem.
3. Implement a solution algorithm to the relaxed problem.
4. Perform numerical experiments to analyze whether the developed formulationhelps recover the solution to the original power flow problem.
Prerequisites The student should be familiar with mathematical optimization theoryand techniques.
P LPower
Systems
Laboratory
Decentralized transmission capacity allocationfor the exchange of energy and reservesSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Type Semester/Master
Background The appeal for improved efficiency has motivated the establishmentof a single European day-ahead electricity market. However, the integration of thereserve capacity and balancing markets is a necessary in order to reduce the balancingcosts. Given though that these markets are operated by national TSOs, the transmis-sion allocation between energy and reserves to enable the cross-border exchange ofregulating services has to be performed in coordinated but decentralized fashion.
Description This project will develop a decentralized solution algorithm to deter-mine the optimal regional reserve requirements as well as the shares of cross-bordercapacity that need to be allocated to the exchange of reserve services. This modelwill be based on the preemptive transmission allocation model proposed in [1] thatis formulated as a stochastic bi-level optimization problem. The tasks of this work are
1. Familiarize yourself with stochastic optimization, bi-level programming and La-grange decomposition (ADMM).
2. Formulate a decentralized solution algorithm for the preemptive transmission al-location model presented in [1].
3. Evaluate the quality of solution (optimality gap) and the computational perfor-mance of the decentralized algorithm.
4. Implement a cost sharing mechanism (e.g. Shapley value) for the allocation ofcosts/benefits among the different TSOs (depending on time availability).
[1] S. Delikaraoglou and P. Pinson “Optimal allocation of HVDC interconnections for exchangeof energy and reserve capacity services” in Energy Systems (2018): 1-41.[2] A. Ahmadi-Khatir, A. Conejo, and R. Cherkaoui , “Multi-area unit scheduling and reserve allo-cation under wind power uncertainty”, in IEEE Transactions on Power Systems, 29.4 (2014).
Prerequisites Interested students should have a strong background on optimizationand power systems operation. A good understanding of applied statistics is an advan-tage. The optimization models may be implemented in Python or Matlab dependingon the background of the student. Given the novelty of this research topic, there isencouragement from our side for turning the student project into a published pa-per. Students who are capable and willing to go the extra mile and get their workpublished (with the help of the supervisors) are preferred.
P LPower
Systems
Laboratory
BSc Gruppenarbeit: Development of a databaseenvironment for power systems simulationsSupervisor Adrian Hauswirth, [email protected]
Type Group Project (2-3 people)
Background Research in power systems relies heavily on large-scale numerical sim-ulations and data analysis. However, today’s standard procedures in academic researchfor handling power systems data are outdated, tedious and often rely on closed-sourcesoftware such as MATLAB.
Description
In this project, we want to rethink the handling of power systems research data. Thegoal is to develop a database environment based on open-source tools that can re-place and unify current file-based approaches. In other words, we want to define acommon interface for power system data storage, simulations, and analysis.
The project will follow an iterative development process with multiple milestones:
1. A first step will consist in figuring out appropriate database schemas and codingscripts to import/export to a power system data format called matpower.
2. In the further steps, we will modify our existing simulation framework to integratewith the new database. We also want to interface a recently-developed visualiza-tion framework (https://git.ee.ethz.ch/jhuwyler/visual-net/).
3. Finally, a key requirement is to document everything such that the final code canbe made publicly available.
Sounds boring! Why bother? This project is a great opportunity to do some high-quality software development up to the point where the code can be published online.Students can gain valuable experience in database design and data handling.
Finally, students will learn a lot about power systems modeling, but with as lit-tle math as necessary. Hence, the project can also serve as preparation for futuresemester/master projects in this domain.
Prerequisites Students should have very good general coding skills and be able learnnew programming languages/technologies quickly and independently (expect the useof MATLAB, SQL, JSON, Javascript, Python, ...). Dedication to clean code and excellentdocumentation is vital. Exposure to power systems modeling is beneficial, but notnecessary.
P LPower
Systems
Laboratory
Fault identification in distribution gridsSupervisor Stavros Karagiannopoulos, [email protected]
Dr. Omid Alizadeh-Mousavi (DEPsys), [email protected]
Type Semester/Master
Background DEPsys is a Swiss energy company focused on integrating decentral-ized renewable generation (e.g., solar photovoltaic) and distributed storage into low-voltage electricity grids, and its GridEye software helps distribution system operatorsmonitor and optimize their networks. The identification and localization of faultsplays an important role in the secure operation of distribution grids, and increasingamounts of distributed energy resources has increased the necessity and difficulty ofaccurate fault detection and identification. This project uses existing GridEye mod-ules to develop new algorithms to identify and localize different kinds of faults usingvoltage and current measurements provided by GridEye.
Description This project will develop and implement new approaches for faultidentification in distribution networks by using voltage and current measurementsin low-voltage grids and current measurements in medium-voltage grids to identifyand localize different types of faults. Tasks are to
1. Study faults in distribution grids considering distributed energy resources, differenttypes of neutral grounding, and unbalanced three-phase loading.
2. Develop algorithms for fault identification and localization in low-voltage andmedium-voltage grids (using low-voltage grid measurements).
3. Validate these algorithms in simulation software (e.g. EMTP-RV) and/or using areal-time simulator.
4. Implement the developed methods in the GridEye platform.
Prerequisites The electrical engineering student should have some familiarity withelectric power systems and three-phase distribution networks; experience writingefficient and reusable code in C, Python, and MATLAB; experience working with realdata; good knowledge of power system protection and fault identification; and somefamiliarity with power system simulation software and real-time system simulators.
P LPower
Systems
Laboratory
HVDC transmission investments using FlowgateTransmission RightsSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Type Semester/Master
Background In the forthcoming years, a significant share of new interconnectionsin the European power system will be based on High Voltage Direct Current (HVDC)technology. In a regulated environment, transmission expansion is decided based onsocial welfare maximization. On the other hand, in a liberalized market framework,transmission investments are carried out by private investors and thus proper mer-chant mechanisms must be developed to support the new transmission infrastructure.
Description This project will investigate new merchant mechanisms for electricitytransmission investments based on DC Flowgate Transmission Rights (DC-FGR). Thegoal of the project is to assess whether DC-FRGs and locational prices provide suf-ficient incentives for long-term investment in the transmission network. In addition,merchant model solution will be compared against a regulated transmission expan-sion model to investigate its effect on social welfare. The tasks of this project are
1. Literature review on transmission investment models and FGR financial tools.2. Formulation of a stochastic HVDC transmission expansion model that maximizes
expected social welfare.3. Formulation of a merchant HVDC transmission investment model that maximizes
investment profit through the collection of DC-FRGs.4. Evaluate and compare the transmission investment plants from steps 2 & 3.5. Application of decomposition techniques for the implementation of models 2 & 3
on real-life power systems (depending on time availability).
[1] S. Delikaraoglou, J. Taylor, P. Pinson “Direct Current Flowgate Transmission Rights”[2] T.Kristiansen and J.Rosellon , “A Merchant Mechanism for Electricity Transmission Expansion”,Chapter 6 in Financial Transmission Rights, Springer, London, 2013.[3] Conejo, Antonio J., et al. “”Investment in electricity generation and transmission.” DecisionMaking Under Uncertainty. Springer, New York, 2016.
Prerequisites Interested students should have a strong background on optimization andpower systems operation. A good understanding of applied statistics is an advantage. Theoptimization models may be implemented in Python or Matlab depending on the backgroundof the student. Given the novelty of this research topic, there is encouragement from our sidefor turning the student project into a published paper. Students who are capable and willing togo the extra mile and get their work published (with the help of the supervisors) are preferred.
P LPower
Systems
Laboratory
Integrating hydropower operations into asimulation of the Swiss power systemSupervisor Dr. Jared Garrison (FEN, ETH Zurich), [email protected]
Dr. Turhan Demiray (FEN, ETH Zurich), [email protected]
Type Master
Background Hydropower operations have longer planning horizons than other types of generation: plants are optimized to store and release water during the year such that generation occurs during months with higher electricity prices. Since conven-tional power system models only simulate small parts of the year, they do not capture this long-term behavior. Additionally, hydrological models with time resolutions capa-ble of modeling hourly water inflows are computationally complex and do not model the power system objectives of hydropower operations. This thesis will bridge these research areas. The developed methods methods and inflow data sets will be critical for studying the Swiss power system as hydropower provides most of the countrys electricity generation.
Description Hydropower operations have longer planning horizons than other types of generation: plants are optimized to store and release water during the year such that generation occurs during months with higher electricity prices. Since conven-tional power system models only simulate small parts of the year, they do not capture this long-term behavior. Additionally, hydrological models with time resolutions capa-ble of modeling hourly water inflows are computationally complex and do not model the power system objectives of hydropower operations. This thesis will bridge these research areas. The developed methods methods and inflow data sets will be critical for studying the Swiss power system as hydropower provides most of the countrys electricity generation.
1. Review the hydropower modeling literature.
2. Review and analyze the available hydrological (e.g., inflows) and hydropower data (e.g.,plant information).
3. Develop computationally-efficient models of hydropower operations and evaluate themagainst a one-year baseline model.
4. Calibrate the selected methods against historical hydropower operations.
5. Write the report and present the results to Prof. Dr. Paolo Burlando.
Prerequisites The student should have expertise or experience in both power system simulation and hydrological systems operation. Experience in Matlab is required, and experience using MatPower is a plus. This thesis will be jointly supervised by members of the Research Center for Energy Networks (FEN) and the Department of Hydrology and Water Resources Management.
P LPower
Systems
Laboratory
BSc Gruppenarbeit: Interactive browser-basedvisualization for AC power flow solutionsSupervisor Adrian Hauswirth, [email protected]
Type Group Project (2-3 people)
Background The study of power flows in electricity grids is often facilitated by mean-ingful visualizations. In particular, the development of powerful web-based visualiza-tion frameworks for data analysis offer new possibilities for this purpose.
In this project, we want to reproduce the results in the classic paper [1] and use asand dance visualization to interactively present the data. For examples see http://stardustjs.github.io/examples/sanddance/.
This type of visualization will in particular support the presentation of our on-goingresearch on real-time power system optimization [2].
Description A first task for the project is to understand the problem of load flowcomputations and the basic ideas behind the methods used in [1]. From there on, theproject has two main parts:
1. A first part consists in reproducing the data in [1]. This requires the configurationand use of a continuation power flow solver. For this part, good MATLAB coding skillsare required.
2. The second part consists in setting up the interactive visualization. This requiressome experience with web development (Javascript, HTML, CSS).
Further steps (for a 3-person team) could include a high-performance implementationof the power flow solver that can solve for new data points in close to real-time.
Finally, an important part of the project is to make the everything ready for onlinepublication, e.g., as a website. This implies a clean user interface and documentation.
[1] I. A. Hiskens and R. J. Davy, “Exploring the power flow solution space boundary,” in IEEETransactions on Power Systems, vol. 16, no. 3, pp. 389-395, Aug. 2001.[2] A. Hauswirth et al., “Online optimization in closed loop on the power flow manifold,” inPowerTech 2017.
Prerequisites The students should have good general coding skills, preferably inMATLAB and some experience with web development. A good background in numer-ical methods is helpful. Exposure to power systems modeling is beneficial, but notrequired.
P LPower
Systems
Laboratory
Multi-carrier energy systemsSupervisor Conor O’Malley, [email protected]
Type Semester/Master
Background The contemporary energy sector comprises mainly electrical, naturalgas, and district heating utilities. Traditionally, these utilities have strove to design andoperate their networks in an optimal way to ensure an efficient and reliable energysupply to consumers. More recently, environmental concerns have led to an increasedpenetration of renewable energy resources in these networks. The integration of highpenetrations of renewable energy can be challenging for each of these systems indi-vidually due to their inherent seasonality and variability. One solution would be to relyon other energy sectors for energy regulation. This approach requires holistic analysisof these systems to assess their interdependencies and facilitate energy interactions.
Description Master and semester projects can be carried out in areas like thosebelow. If you are interested in any of these topics or other related ones, get in touch!We should be able to define an interesting project.
• Develop integrated models of electricity, gas, and heating networks at the trans-mission and distribution level.
• Investigate how flexibility between different energy carriers can be used to inte-grate renewables, defer grid investment, and increase network reliability.
• Study the impact of Combined Heat and Power (CHP) plants on future energy net-works with high levels of renewable energy resources.
P LPower
Systems
Laboratory
Optimal power flow formulations in moderndistribution gridsSupervisor Stavros Karagiannopoulos, [email protected]
Dmitry Shchetinin, [email protected]
Type Master/Semester
Background In the future, active distribution networks are expected to host a largeamount of renewable energy sources (RES) that will supply a growing share of totaldemand. These units, in combination with other distributed energy resources (DERs),energy storage systems, and flexible loads, will amplify the role of distribution grids.In order to utilize the operational flexibility of such DERs, optimal power flow calcu-lations for distribution grids are gaining more attention.
Description This thesis is motivated by recent advances in computational algo-rithms and increased penetration of DERs in distribution grids. In the context of thisthesis, the student is expected to compare known power flow and optimal power flowformulations and solution techniques in terms of robustness, computational speed,and applicability. The following versions of AC OPF for distribution grids will be inves-tigated: standard AC OPF, AC OPF based on backward/forward sweep (BFS) approach,and convex relaxations of AC OPF. MATLAB code for the first two formulations will beprovided by the supervisors. If time permits, the student could also investigate de-composition techniques for solving AC OPF in distribution grids. (Note that the listedtasks are suggestions and the project can be modified on the fly according to whatthe student finds to be the most interesting topic to explore.) Tasks are
1. Literature survey, including optimal power flow for distribution grids (convex re-laxations, decomposition techniques, etc.).
2. Understand the implementation of the full AC OPF and the BFS OPF problems.
3. Formulate and implement the full AC optimal power flow problem with convexrelaxations.
4. Formulate and implement decomposition of AC OPF for distribution grids.
5. Run the different OPF calculations using several distribution test systems and com-pare the methods based on robustness, computational speed and applicability.
Prerequisites The student should be familiar with Python and/or MATLAB and havesome knowledge of mathematical optimization theory and techniques.
P LPower
Systems
Laboratory
Optimization of First Zone Distance Protectionfor Flexible Transmission LinesSupervisor Nadezhda Davydova, [email protected]
Dmitry Shchetinin, [email protected]
Type Semester/Master
Background Protection systems are indispensable to ensure safe operation of thegrid. The main goal of any protection system is to localize internal faults while not re-acting to external ones. In distance protection, which is the most widely used type ofprotection in transmission grids, this is achieved by constructing tripping boundariesof protection zones. Traditionally, these boundaries are selected based on physicalconsiderations and engineering experience. This can lead to the protection zonesbeing overly conservative, which prevents the high-speed first zone of the protectionsystem from detecting certain internal faults that can potentially be distinguishedfrom external ones. This is undesirable because the delayed tripping of faulted linesmay cause unnecessary disconnection of distributed generators and stability prob-lems in the grid. The installation of series compensation devices further reduces theprotection zones.
Description The main objective of this thesis is to extend and test a recently devel-oped in PSL optimization-based algorithm for constructing the tripping boundary ofthe first protection zone for transmission lines with series compensation devices (SCs).The algorithm has to account for measurement uncertainties, variations in line load-ings, various possible locations of series compensation devices, and precise modelsof the compensation devices. The exact topic can be tailored to the interests of thestudent. Potential tasks for the project include but are not limited to the following:
1. Review distance protection algorithms for lines with SCs and familiarize yourselfwith the algorithm developed in PSL.
2. Incorporate measurement uncertainties into the optimization problem.
3. Extend the boundary construction algorithm to be applicable for lines with variouslocations of compensation devices.
4. Consider grid loading in optimization problem.
5. Test the developed algorithm for various fault instances and normal grid operationmodes.
Prerequisites The student should be familiar with MATLAB.
P LPower
Systems
Laboratory
Optimized decentralized control schemes foractive distribution grids based on ML approachesSupervisor Stavros Karagiannopoulos, [email protected], PSL
Prof. Duncan Callaway, [email protected] UC Berkeley
Type Semester/Master
Background In order to ensure secure system operation at a low cost, centralizedand decentralized operational schemes are used to optimally dispatch these units.This thesis will explore decentralized, real-time, operation schemes for the optimaldispatch of distributed energy resources in absence of extensive monitoring and com-munication infrastructure.
Description The aim of this thesis is to investigate various optimized decentral-ized approached and assess their suitability to cope with current grid challenges. Theexamined methods will be based on machine learning [1, 2], derivation of individ-ual inverter characteristic curves [3], and other voltage droop control schemes [4] indistribution grids.
1. Review operation control schemes of active distribution grids and familiarize your-self with Optimal Power Flow schemes.
2. Model specific functional forms of decentralized control responses found in theliterature, e.g. [4]
3. Extend these methods with more flexible functional forms (e.g. smoothing splines)and quantify the benefits of adding degrees of freedom in the model that is usedto describe the local droop control
4. Demonstrate the derived operation methods through case studies and comparethe operation stage of the proposed method against current schemes in existinggrid codes (e.g. German and Italian)
[1] O.Sondermeijer,et al., ”Regression-based inverter control for decentralized optimalpower flow and voltage regulation,” in Proceedings of the IEEE PES GM, 2016.[2] F. Bellizio, et al., ”Optimized local control schemes for active distribution gridsusing machine learning techniques”, IEEE PES General Meeting 2018.[3] S. Karagiannopoulos, et al., ”Hybrid approach for planning and operating activedistribution grids,” IET GTD, vol. 11, no. 3, p. 685 695, 2017.[4] K. Baker, et al., ”Network-Cognizant Voltage Droop Control for Distribution Grids”,IEEE Transactions on Power Systems, DOI: 10.1109/TPWRS.2017.2735379
Prerequisites The student should be familiar with MATLAB, optimization and ma-chine learning techniques.
P LPower
Systems
Laboratory
Participation in grid optimization competitionorganized by ARPA-ESupervisor Dmitry Shchetinin, [email protected]
Dr. Tomas Tinoco De Rubira (EPRI)
Type Semester
Background The application of optimization algorithms to power system relatedproblems has attracted significant attention from the research community due to itshigh practical importance. However, the complexity of power systems models makeoptimization in power systems difficult. Despite significant efforts, to this day thereis no benchmark algorithm for solving the AC Optimal Power Flow (OPF) problem. Toaccelerate the development of transformational and disruptive methods for solvingpower system optimization problems, the Advanced Research Projects Agency-Energy(ARPA-E) in the United States has recently begun the Grid Optimization Competition.Anyone can participate in this competition, the goal of which is to solve the SecurityConstrained OPF problem in the best and fastest way possible. Details about the com-petition including rules, scoring, timeline, prizes, etc. are available on the competitionwebsite.
Description The goal of this project is to participate in the grid optimization com-petition organized by ARPA-E. In order to do this, the student will develop interfaceroutines for interacting with the competition website. This includes data parsers, out-put formatting, and so on. All relevant information is available on the competitionwebsite. The student then will use several algorithms provided by the supervisors toparticipate in the competition. Depending on the interest of the student, he/she canalso work on tuning the algorithms to improve the obtained score. Tasks are to
1. Familiarize oneself with competition rules and requirements.
2. Implement necessary interface routines for using the competition website.
3. Participate in the competition by trying out several (provided) solution algorithms.
Notes The competition website is https://gocompetition.energy.gov/.
P LPower
Systems
Laboratory
Planning and operating schemes for activedistribution gridsSupervisor Stavros Karagiannopoulos, [email protected]
Dr. Petros Aristidou (Univ. of Leeds), [email protected]
Type Semester/Master
Background In the future, active distribution networks are expected to host a largeamount of renewable energy sources (RES) that will supply a growing share of totaldemand. These units, in combination with other distributed energy resources (DERs),energy storage systems, and flexible loads, will amplify the role of distribution gridsand permit them to provide power and ancillary services to higher voltage levels.However, such developments cause significant planning and operating challenges.
Description This project will investigate new approaches for planning and oper-ating active distribution grids. In the planning stage, a mixed-integer mathematicaloptimization framework will be extended from [1]. For operations, central and localcontrol schemes will be developed based on [2] and [3] and considering unbalancedloading and uncertainty. The product of this thesis will be a decision tool for DSOs toassist them in system planning and operation. Tasks are to
1. Review the operational planning and real-time control of active distribution grids.2. Familiarize yourself with linear and nonlinear optimal power flow and with the
unbalanced operation of multi-phase networks.3. Compare solutions obtained using full ACOPF and the iterative Backward Forward
Sweep method [1,3] in terms of convergence and optimality.4. Evaluate developed planning and operation methods using both a benchmark ra-
dial grid and part of the distribution grid of EWZ having a meshed configuration.5. Compare the operation stage of the proposed method against schemes in existing
grid codes (e.g., Germany and Italy).
[1] S. Karagiannopoulos, P. Aristidou and G. Hug, “Co-optimisation of planning and operation foractive distribution grids”, in PowerTech 2017.[2] S. Karagiannopoulos, P. Aristidou, G. Hug, “Hybrid approach for planning and operating activedistribution grids”, IET Generation, Transmission and Distribution, 11 (3): 685-695, 2017.[3] S. Karagiannopoulos, L. Roald, P. Aristidou and G. Hug, “Operational planning of active distri-bution grids under uncertainty”, in IREP 2017.
Prerequisites The student should be familiar with MATLAB and mathematical opti-mization techniques.
P LPower
Systems
Laboratory
Power flow solution algorithm for detectingsources of infeasibility in the systemSupervisor Dmitry Shchetinin, [email protected]
Dr. Tomas Tinoco De Rubira (EPRI)
Type Semester/Master
Background Steady-state power flow analysis is a basis of power system planningand operation. Reliable power flow solution algorithms are required to operate systemsecurely and efficiently. With higher penetrations of renewable generation, the flowsin the system are becoming increasingly variable and less predictable. This may leadto some credible system states being infeasible, which means that the system cannotsustain a certain load and generation pattern. In such a case, it is imperative to knowwhich loads and/or generators are primarily responsible for this infeasibility. Sincetraditional methods of power flow analysis do not provide such information, newalgorithms have to be developed to increase the awareness of system operators andhelp them decide on their actions.
Description The main goal of this project is to develop a new algorithm for solvingthe power flow problem, which works reliably for feasible cases and provides infor-mation regarding the main sources of infeasibility for infeasible ones. To achieve thisgoal, the power flow problem should be cast as an optimization problem that mini-mizes the sum of absolute values of node power balance mismatches. Due to suchproblem formulation, typically only a few power balances will be nonzero at the solu-tion for an infeasible case, which will correspond to load/generators that are mainlyresponsible for infeasibility. A number of test cases have to be developed to evaluatethe algorithm’s efficiency and ability to detect the sources of infeasibility. Tasks are to
1. Review literature on relevant solution algorithms.
2. Implement a power flow solution algorithm based on L1-norm minimization.
3. Develop a series of test cases corresponding to infeasible system states.
4. Perform numerical experiments to analyze the algorithm’s ability to detect ‘prob-lematic’ buses in the system.
Prerequisites The student is encouraged to use power flow analysis software avail-able at https://github.com/ttinoco. Although there are no prerequisites for this project,knowledge of Python is an advantage.
P LPower
Systems
Laboratory
Sharing storage recourses using tradablefinancial/physical rightsSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Type Semester/Master
Background Energy storage systems (ESS) may improve power system flexibility andreduce the uncertainty resulting from renewables’ forecast errors. Nonetheless, thelarge-scale integration of ESS in the power system requires new market and regulatoryframeworks to promote efficient operation and incentivize new investments. Recentliterature proposes the definition of new market mechanisms, like physical or financialstorage rights, to enable open access to ESS for all market participants. However, theseapproaches are not mature and further studies are needed to assess their efficiency,especially in markets with large shares of stochastic in-feed.
Description This project will investigate the effect of physical and/or financial stor-age rights on market efficiency (i.e., expected system cost). An auction-based allo-cation model will be used to assess the potential gain or loss from the introductionof this mechanism for market players with different generation portfolios. For thepurpose of this work, we will employ the mathematical framework of stochastic equi-librium models casted as mixed complementarity models. Tasks are to
1. Review the literature and acquire a good level of understanding of physical andfinancial storage rights.
2. Study the mathematical framework of (stochastic) mixed-complementarity models.
3. Formulate a (stochastic) equilibrium model for the allocation of physical/financialstorage rights to market participants.
4. Simulate the market outcome for different power system setups and evaluate theeffect of storage rights on social welfare and on market players’ profits.
Prerequisites Interested students should have a strong background in optimization,and a good understanding of applied statistics is an advantage. Equilibrium modelsmay be implemented in Python, MATLAB, or GAMS depending on the background ofthe student.
P LPower
Systems
Laboratory
Statistical parameterization in MDPC-basedprivacy protectionSupervisor Jun Xing “Jack” Chin, [email protected]
Dr. Tomas Tinoco De Rubira (EPRI)
Type Master
Background With the roll out of Advanced Metering Infrastructure (AMI) withinthe EU region, Utility Providers (UPs) are able to gather much more data about theircustomers at greater resolution. This has led to concerns about consumer privacy dueto inferences that can be made from this data. To protect consumer privacy, an EnergyManagement Unit controller based on Model-Distribution Predictive Control (MDPC)has recently been proposed. While causal and implementable, the proposed schemedhas limited scalability in order to maintain computational tractability. The goal of thisproject is to incorporate more advanced statistical estimation methods, such as theparameterization of the probability distribution functions, into the MDPC scheme.
Description This project will (a) explore statistical models that best represent res-idential consumer load and (b) parameterize selected models and incorporate theminto the MDPC scheme. Tasks are to
1. Review literature about the parameterization of statistical estimates from timeseries data.
2. Formulate optimization problems that balance energy cost and privacy loss usingparameterized load data.
3. Implement and simulate an Energy Management Unit using the proposed opti-mization program.
4. If time permits, validate the proposed scheme on different samples of actual smartmeter data.
J. Chin, T. Tinoco De Rubira, and G. Hug, “Privacy-protecting energy management unit throughmodel-distribution predictive control”. Available: https://arxiv.org/pdf/1612.05120
P LPower
Systems
Laboratory
Strategic virtual bidding and self-schedulingunder imperfect informationSupervisor Dr. Stefanos Delikaraoglou, [email protected]
Prof. Jalal Kazempour, [email protected]
Type Semester/Master
Background Virtual bidding is a form of financial arbitrage performed between theday-ahead and real-time markets by players who do not own any physical assets. Self-scheduling refers to the case in which generators optimize internally their productionschedule, which is then offered to the market at zero price. In theory, these mech-anisms improve the convergence of day-ahead and expected real-time prices andprovide some operational flexibility to the slow-start generators. Nonetheless, thesepositive effects may diminish if the market players act strategically or when their de-cision making is based on imperfect information about the underlying uncertainties.
Description This project will investigate the impact of imperfect information andstrategic behavior of virtual bidders and self-scheduling generators on the marketefficiency. The optimal bidding/offering strategy of virtual bidders/self-schedulingunits will be formulated as a stochastic bi-level problem. The tasks of this work are
1. Perform a literature review on virtual bidding & self-scheduling mechanisms andon bi-level optimization.
2. Formulate the optimal bidding/offering strategy of a virtual bidder/self-schedulingunit as a stochastic bi-level program.
3. Evaluate the impact of strategic behavior and imperfect information of virtualbidders/self-scheduling generators on market efficiency.
[1] Birge, John R., et al. ”Limits to Arbitrage in Electricity Markets: A case study of MISO.” inEnergy Economics (2018).[2] J.Kazempour and B.F.Hobbs. ”Value of Flexible Resources,Virtual Bidding,and Self-Schedulingin Two-Settlement Electricity Markets with Wind GenerationPart I: Principles and CompetitiveModel.” in IEEE Transactions on Power Systems 33.1 (2018).
Prerequisites Interested students should have a strong background on optimization andpower systems operation. A good understanding of applied statistics is an advantage. Theoptimization models may be implemented in Python or Matlab depending on the backgroundof the student. Given the novelty of this research topic, there is encouragement from our sidefor turning the student project into a published paper. Students who are capable and willing togo the extra mile and get their work published (with the help of the supervisors) are preferred.
P LPower
Systems
Laboratory
BSc Gruppenarbeit: The Grid Game 2.0Supervisor Adrian Hauswirth, [email protected]
Type Group Project (3-4 people)
Background Unleash your inner game programmer! The objective of this project isto develop a browser game for the public that illustrates the challenges that powergrid planners are facing today.
The Energiewende has fundamental consequences for our electric power supply: Cen-tralized bulk power plants are gradually replaced with volatile and decentralizedenergy sources such as photovoltaics and wind turbines. This makes investmentsdecisions into power grid infrastructure very challenging.
To highlight these problems, we want to develop a browser game that serves as aneducational, yet entertaining tool to inform the public.
Description We envision a round-based browser game (HTML5) that requires play-ers to make investment decisions every round (similar to a Tower Defense game). Forinstance, at every round the player has to decide whether to invest in additional gen-erators, power lines, storage devices, etc. As the round plays out, actual power flowsare computed and the player can see whether with the new investments, the powergrid works well and whether all power demand is satisfied.
The project will follow an iterative development process with multiple milestones.
1. A first stage will consist in setting up a base game with limited features. Namely,the base game should focus on distribution grids and the simple trade-off betweensolar plants, line upgrades, and storage. At the end of this stage, several small levelsshould be designed and balanced.
2. A second feature set could incorporate voltage problems in distribution systems aswell as smart devices.
3. Further milestones could include features that are primarily encountered in high-voltage transmission grids, such as meshed grids, ramping constraints, etc.
A key aspect is to make the game ready for publication, hence the graphical design,game balancing and documentation are an integral part of the project.
Prerequisites The students should have a good programming background, preferablywith experience in web and/or game development. A knack for nice user interfacesand clean visual presentation is also highly welcome. Exposure to power systemsanalysis is beneficial, but not required.
P LPower
Systems
Laboratory
Traveling Wave Protection for Distribution Gridswith Renewable Energy SourcesSupervisor Nadezhda Davydova, [email protected]
Type Semester/Master
Background High penetrations of distributed generation (DGs) in distribution gridsand development of microgrids may cause malfunctioning of conventional distribu-tion level protection systems. Despite multiple works dedicated to addressing thisproblem, the development of reliable, high-speed, and cost-efficient protection sys-tems for active grids remains a topical issue. Traveling wave based protection systemsis a promising solution since they are fast and essentially not affected by the operationmodes of DGs and microgrids.
Description The main objective of this thesis is to extend and test a recently devel-oped in PSL traveling wave protection algorithm [1] for distribution grids with DGs.The exact topic can be tailored to the interests of the student. Potential tasks for theproject include but are not limited to the following:
1. Review traveling wave theory and its application in protection algorithms.
2. Make the protection algorithm robust to switching transients and lightning strikes.
3. Extend the protection algorithm to be applicable for lines with various sources ofdiscontinuities.
4. Add model of surge arresters to the test grid in EMTP-RV.
5. Test the developed algorithm for various fault instances and normal grid operationmodes.
[1] N. Davydova and G. Hug, ”Traveling Wave Based Protection for Medium VoltageGrids with Distributed Generation”, in PowerTech 2017.[2] N. Davydova and G. Hug, ”Wavefront-based Protection for Active Distribution Grids”,in ISGT Europe 2017.
Prerequisites The student should be familiar with MATLAB.
P LPower
Systems
Laboratory
Unit Commitment Problem with Inclusion ofInertia ConstraintsSupervisor Uros Markovic, [email protected]
Dr. Stefanos Delikaraoglou, [email protected]
Type Semester/Master
Background Inverter-connected wind turbines and solar photovoltaics do not pro-vide rotational inertia and displace synchronous machines. Traditional assumptionsregarding grid inertia are thus invalid for systems with lots of renewables. This hasimplications on frequency dynamics and power system stability and operation, sincethe dynamics are faster in power systems with less inertia. In order to secure fre-quency stability, system operators might prescribe the minimum inertia requirementsthat must be provided at all times by the available units. However, such requirementsmight affect the traditional unit commitment problem, as they could enforce the ac-tivation of unscheduled generating units.
Description This project will investigate the unit commitment problem in powersystems with critically low levels of rotational inertia. The additional stability crite-ria will be determined from the swing equation and incorporated as critical inertiaconstraints. Furthermore, stochastic and/or robust optimization techniques will beincorporated, together with the real-time dispatch controllers. Tasks are
1. Review the literature on unit commitment in low inertia inertia power systems.
2. Understand and extend the existing unit commitment problem in Python.
3. Derive and include the critical inertia constraints into the traditional unit commit-ment problem.
4. Incorporate uncertainty for maximum power imbalance (N−1 reliability), stochasticnature of renewables and forecast errors.
5. Improve real-time dispatch of units providing inertial response (continuous unitcommitment and/or flexible ramping).
Prerequisites Interested students should have a basic understanding of power sys-tems analysis. Experience with Python and a strong mathematical background isadvantageous. Since research into low- and no-inertia power systems is immature,there is encouragement from our side for turning the student project into a publishedpaper. Students who are capable and willing to go the extra mile and get their workpublished (with the help of the supervisors) are preferred. This thesis is within theframework of the MIGRATE project: https://www.h2020-migrate.eu/.