Jose Marti

Traditional power systems have high inertia, and disturbances can be represented with relatively large discretization steps. However, power electronic interfaced generation requires time steps, at least two orders of magnitude smaller to represent the switching operations. In addition, the increased penetration of these devices reduces the total inertia of the system, and a single-time step size transient stability (TS) solution that considers the details of the power electronic devices would take two orders of magnitude longer (that is, electromagnetic transients (EMT) simulators time step sizes). Hybrid TS-EMT simulators aim at using large time steps for the traditional (slow) part of the system and small time steps for the power electronic devices. However, in the current state-of-the-art, the proposed hybrid solutions are limited in a number of ways. Most of them introduce a one TS time step interfacing delay between solutions, which reduces accuracy and can create numerical oscillations. Other approaches use the travelling time of transmission lines to tear the system and interface the solutions, but this limits the flexibility of connecting power electronic components anywhere in the system.The hybrid solution presented in this work combines a Shifted Frequency Analysis (SFA) solution that is more accurate than traditional phasor solutions to follow transient stability trajectories in lower inertia systems with the traditional EMT solution for the fast power electronic converters. In order to interface the SFA and EMT solutions, a new hybrid protocol has been developed that overcomes the disadvantages of the current hybrid solutions. The two subsystems are joined using the Multi-Area Thévenin Equivalent (MATE) solution framework, and introducing the novel concept of employing a second parallel EMT simulation. Illustrative examples demonstrate the validity and accuracy of the SFA-EMT multirate protocol. The new protocol achieves significant computational savings when compared with an all-EMT solution. A simulator operating with the proposed protocol will allow improved accuracy when conducting power system transient studies with high penetration of renewable generation without considerably increasing the computational time.

View record

This dissertation aims to present a full-stack concept design for an earthquake early warning platform utilizing a dense array of sensors embedded in consumer electronic devices. The proposed design includes four systems: observation, estimation, prediction, and decision systems, each with a specific type of intelligence.A novel change detection technique, specially devised for the observation system, is responsible for detecting and estimating the arrival time. Knowing the geographical location and arrival times collected from various sensors allows the platform to compute an approximate location of the event through an optimization method. The performance results on 732 ground motions indicate that the error in arrival time estimation is less than 1.5 ?, on average. Meanwhile, the event location estimation average error is less than 16 ??, with a 99% confidence level.Furthermore, a novel earthquake intensity prediction model based on a neural network structure is responsible for evaluating the size of the event. The proposed prediction model yields 57% standard error over 4691 carefully selected ground motions. In contrast, the performance results of the voting process responsible for updating the alarm status indicate the overall accuracy of 75% for the platform; that is, three-quarters of the ground motions are correctly classified, on average. However, the average false alarm rates are 20% to 30%. Accordingly, investigating the effect of the threshold value on the false alarm rates illustrates that consideration of the practicality in the design concept allows the selection of an optimal alarm threshold.Utilizing the Internet of Things (IoT) infrastructure for earthquake early warning is a compelling and challenging alternative compared to conventional platforms. Future works require collaboration between academia and industry to devise guidelines for reconfiguring IoT infrastructure and improving the performance of individual earthquake early warning systems.

View record

Modern societies depend on the proper and resilient functioning of their critical infrastructures (CIs) to support quality of life to their population. The interdependencies of the CI systems, however, make the CIs increasingly vulnerable to several threats; as a result, natural or human-made disasters can cause significant physical, economic, and social disruptions. Since total beforehand protection cannot be guaranteed, CI protection strategies should focus on resilience enhancement at both the pre- and the post-disaster phase for a better response. This thesis first proposes a resilience assessment framework that provides quantitative means to assess infrastructure resilience for interdependent CI systems. The framework is tested within the Infrastructure Interdependencies Simulator (i2SIM) that models and simulates the CI interdependencies. A coupled Cyber-Physical System (CPS) is modelled within the i2SIM framework to study the process of cascading failures. Resilience enhancement for the pre-disaster preparedness is done with a risk evaluation approach and is proposed by considering the profile of a hurricane to calculate the probability of failure of the network components. Based on the results of the evaluation, a resource allocation optimization is formulated using mixed-integer nonlinear programming (MINLP). For post-disaster resilience enhancement, an optimal reconfiguration algorithm, together with a hybrid load shedding strategy, is developed to find alternative paths to maintain supply to the most critical loads. A modified shortest path search that we call the Optimal Recovery Sequencing (ORS) is used to optimize the repair sequences. The obtained numerical results validate that the recovery ability of the coupled system, and as a result its resilience, is increased with the proposed optimization. Finally, to reduce the computational complexity for very large scenarios and for real-time response, this thesis uses a Soft-Hard Optimal Convergence (SHOC) methodology. Machine learning techniques are used to train an artificial intelligence (AI) agent with thousands of off-line scenarios to provide the initial estimate for the SHOC algorithm. After the disaster occurs, hard optimization methods are used on a small subset of solutions identified by the AI agent. Using this methodology, solution time speedups of 800 times for a 70-bus test system with 30 simultaneous faults are achieved.

View record

In this thesis we present the implementation of a coordinated decision-making agent for emergency response scenarios. The agent’s implementation uses Reinforcement Learning (RL). RL is a machine learning technique that enables an agent to learn from experimenting. The agent’s learning is based on rewards, feedback signals proportional to how good its actions are. The simulation platform used was i2Sim, the Infrastructure Interdependencies Simulator, in which, we have tested the suitability of the approach in previous studies. In this work, we have added new features, for increasing the speed of convergence and enabling distributed processing capabilities. These additions include enhanced reward and exploration schemes and a scheduler for orchestrating the distributed training. We include two test cases. The first case is a compact model with 4 critical infrastructures. In this model, the agent’s training required only 10% of the attempts as compared to references given by past studies done in our group. Improvements in convergence come from the enhanced shaping reward and exploration schemes. We trained the agent across 24 simultaneous configurations of our model (scenarios). The complete distributing training process needed 4 minutes. The second case is an extended model, a more detailed representation of the first case. This extended case included additional infrastructures and a higher level of resolution. By adding more infrastructures, the dimensionality of the problem grew four thousand times. This dimensionality growth did not affect performance and the training had an even faster convergence. We ran 96 parallel instances of the extended model and the process completed in 2.87 minutes. The results show a fast and stable convergence framework with a wide range of applicability. This agent could help during multiple stages of emergency response including real time situations.

View record

This research contributes to developing a time domain and a frequency domain formulationsto solve electromagnetic transients in power system with multiconductoroverhead transmission lines.The time domain solution introduces a frequency dependent transmission linemodel “FDLM”. For the development of the FDLM a fundamental constraint isadded to the classical line equations to maintain the symmetry between electric andmagnetic fields. As a result, voltage waves and current waves travel together andthe characteristic impedance remains uniform along the line. With this premise, aconstant real transformation matrix can be obtained to diagonalize the line functionswith high accuracy. This feature can greatly facilitate the line modelling asopposed to the existing line models which require complex frequency dependenttransformation matrices for their diagonalization. The use of a single constant realtransformation matrix for the voltage and current waves which is exact over the frequencyrange enables FDLM to provide higher accuracy and numerical efficiencythan the existing line models while it complies with the physical system.The accuracy of the FDLM is assessed through comparisons with a newly developedDiscrete Time Fourier Series frequency domain solution. This methodologyis based on the correct specification of the time window and frequency windowwidths. Guidelines are provided for this set up which avoids the typical Gibbs andaliasing errors related to the classical frequency domain solutions. The proposedfrequency domain solution is simpler to implement than the most commonly usednumerical Laplace transform solution while it does not require further considerationsto use damping factors or windowing functions.

View record

In recent years, extreme events, such as hurricanes, earthquakes, floods and fires, occur more frequently and at a higher intensity. The growing complexity and interdependence of modern infrastructure systems makes them vulnerable to such events. Emergency response is the process of implementing appropriate actions to reduce human and economic losses following these events. Efficient response requires an understanding of the existing infrastructure systems and their interdependencies. In this thesis, we propose a decision support system for helping emergency responders in making efficient decisions during extreme events. Fires are chosen as an example of the extreme events and firefighting operations as the emergency response to these events. Everyday, fire managers are faced with making increasingly complex manpower decisions; trying to minimize costs and risk levels. The effectiveness of firefighting operations is crucial in minimizing both cost of suppression and economic losses. The contributions of this thesis focus on two levels of fire management plans: operational and strategic. We first develop a methodology to optimize the allocation process of firefighting resources in multiple-fire incidents. The developed methodology employs reinforcement learning, a machine learning algorithm that optimizes the allocation of firefighting units to minimize the total economic losses in the long run. To consider the concept of infrastructure interdependencies in evaluating the economic impact of the incidents, we model a large petrochemical complex using the Infrastructure Interdependency Simulator (i2Sim). In addition, a capacity planning methodology is developed to investigate the impact of manpower investment on the effectiveness of firefighting operations. The developed methodology aims at finding the optimal number of firefighters to be recruited to contain fires and effectively extinguish them. It performs an economic analysis to evaluate the efficiency fire management plans. Finally, we propose a methodology to evaluate the effectiveness of emergency response plans in improving infrastructure resilience. This methodology focuses on two dimensions of resilience: resourcefulness and rapidity. These dimensions are measured by the optimality of allocating firefighting units and by minimizing economic losses. The proposed methodologies are tested using a case study of a large petrochemical complex and promising results are achieved.

View record

Future smart grids will be operated actively in the presence of distributed generators and topological reconfigurations. Distributed Storage Systems (DSS) will also become a viable solution for balancing the load and the intermittent generation of renewable energy sources. The DSS can also provide the smart grid operator with various other benefi ts including peak load shaving, resilience enhancement, power loss reduction, and arbitrage gain.The active nature of future smart grids calls for an accurate state estimation mechanism to serve as a building block for many operational tasks. To that end, the first part of the present thesis leverages the concept of submodularity to solve the problem of robust meter placement for state estimation in reconfigurable smart grids. Next, the thesis proposes a methodology for optimal planning of DSS in smart grids with high penetration of renewable sources. The presented methodology accounts for various advantages of energy storage in smart grids and seeks the optimal trade-off between the investment cost and the expected discounted reward of DSS installation. Finally, the thesis focuses on the problem of Volt-VAR Optimization (VVO) in activesmart grids. The optimal joint operation of reconfiguration switches, energy storage units, under load tap changers, and shunt capacitors is investigated in the presented VVO methodology. The proposed methodologies in this thesis have been tested on sample distribution systems and their effectiveness is validated using real data of smart meters and renewable energy sources.

View record

In this research, static and dynamic voltage stability of distribution systems are studied, with a focus on systems with distributed generation. A voltage dependent model is used for distribution loads, which converts the static voltage stability analysis into a linear problem and allows for real-time solutions. For static voltage stability studies, an index is defined that contains critical data on load and system characteristics. The static voltage behavior of distribution systems, either with or without DG, is studied by analyzing the P-V curves and the proposed voltage stability index at each node of the system. Besides identifying the weakest buses of the system in terms of voltage profile and voltage stability, this approach also allows the extension of the classical voltage stability solution to the increasingly important case of distributed generation placement - where the system is more likely to face voltage transients. The method can also be used in resilience studies. For dynamic voltage stability studies, the Shifted Frequency Analysis (SFA) method is used to evaluate the system transients and its dynamic voltage behavior during and right after being subjected to a change or disturbance in the system. Various scenarios are discussed, including scenarios in which the voltage transients due to DG cause voltage instabilities in the system. SFA and EMTP solutions are compared with each other, and with the static analysis results. SFA and EMTP results also verify the validity of the proposed voltage stability index. The proposed voltage assessment method facilitates real-time decision making, topological reconfiguration to strengthen voltage stability robustness, and DG placement. Different scenarios and distribution system topologies such as looped or meshed distribution systems, as well as microgrids and autonomous/islanded energy grids can be included in the solution.

View record

Our modern societies are dependent on the functioning of infrastructure systems that support economic prosperity and quality of life. These infrastructure systems face an increasingly set of threats, natural or man-made disasters, that can cause significant physical, economic, and social disruptions. Recent extreme events have shown that total protection can not be accomplished. Therefore, Critical Infrastructure Protection strategies should focus not only on the prevention of these events but also on the response and recovery following them. This shift is realized by the concept of infrastructure resilience. In this thesis, we address the problem of assessing and improving infrastructure resilience. The contributions of this thesis focus on modelling, simulation, and optimization of infrastructure systems with respect to their resilience to extreme events. We first develop a resilience assessment framework for interdependent infrastructure systems. The developed framework provides a quantitative means to assess infrastructure resilience by introducing a generalized resilience index. To account for the inherent complexity due to infrastructure interdependencies, we use the i2Sim framework for modelling and simulating the studied infrastructure. The resilience improvement problem is formulated using the proposed resilience index as a resources allocation optimization problem. The problem aims at finding the best allocation of available resources such as power and water to mitigate the consequences of a disaster. Two solutions algorithm are proposed to solve the problem: the first one uses a simulation-optimization approach based on the Ordinal Optimization theory, and the second one uses a Linear Programming formulation. Results of both algorithms show that infrastructure resilience can be greatly improved by efficient allocations of available resources. In addition, a prioritization methodology is developed to assess decision makers to direct resilience investment to the most important components in the infrastructure. Finally, an optimal power distribution network reconfiguration algorithm is developed to complement the two resources allocation algorithms by solving the technical feasibility problem of the power distribution network. A heuristic computationally inexpensive optimization algorithm is developed based on Graph theory for solving this problem. The proposed algorithms are tested using different test cases and promising results are achieved.

View record

Distribution systems (DS) are the last stage of any large power system, delivering electricity to the end-users. Conventionally, simplicity of DS operation has been a priority over its optimality. However, with the recent advancements in the automation and measurement infrastructures, it is now possible to improve the efficiency of DS operation. In this dissertation, a load modeling procedure is proposed which takes advantage of the data available at the smart meters. An algorithm is proposed to decompose the load at each customer level using the smart meter measurements. The proposed load model represents the voltage dependence of loads according to the load composition. Based on the voltage-dependent load model, a linear power flow formulation is developed for DS analysis. The linear current flow equations are then proposed which calculate the branch flows directly without requiring the nodal voltages. Sensitivity factors in terms of current transfer and branch outage distribution factors are also derived using the linear power flow concept. The advantages of having a set of linear equations describing the system statics are demonstrated in a variety of DS optimization problems, such as topological reconfiguration, capacitor placement, and volt-VAR optimization. Using the linear current flow equations, the mixed-integer nonlinear programming problem of DS reconfiguration is reformulated into a mixed-integer quadratic/linear programming problem, which substantially reduces the computational burden of the nonlinear combinatorial problem. Besides developing a direct mathematical optimization approach, a fast heuristic method is also developed here for the minimum-loss network reconfiguration based on the minimum spanning tree problem. This heuristic method provides a good suboptimal solution to initialize the direct mathematical optimization approaches such as branch-and-cut algorithm used for solving combinatorial problems. Based on planar graph theory, an efficient mathematical formulation for the representation of the radiality constraint in reconfiguration problems is introduced. It is shown that this formulation is advantageous over the available methods in terms of accuracy and computational efficiency. The proposed algorithms are tested using a variety of DS benchmarks and promising results are achieved.

View record

National safety and homeland security of an urban community rely heavily on a system of interconnected critical infrastructures (CI’s). The interdependencies in such complex systems give rise to vulnerabilities, which must be accounted for by a proper disaster management. It is a proactive step that is needed to address and mitigate any major interruption in a timely manner. Only then will the management of CI’s be able to appropriately reallocate and distribute the available scarce resources of an existing interdependent system.In this research, we propose an intelligent decision making system that optimizes the allocation of the available resources following an infrastructure disruption. The novelty of our suggested model is based upon the application of a well-known Machine Learning (ML) technique called Reinforcement Learning (RL). This learning method is capable of using experience from a massive number of simulations to discover underlying statistical regularities. Two alternative approaches to intelligent decision-making are studied, learning by Temporal Differences (TD) and Monte Carlo (MC) based estimation. The learning paradigms are explored within the context of competing designs composed of simulators and learning agents architected either independently or together. The results indicate the best learning performance is obtained using MC within a homogeneous system. The goal here is to maximize the number of discharged patients from emergency units by intelligently utilizing the existing limited resources. We show that such a learning agent, through interactions with an environment of simulated catastrophic scenarios (i2Sim-infrastrucutre interdependency simulator), is capable of making informed decisions in a reasonable time.

View record

This thesis presents the development of a novel, transient wave propagation simulator using time-decoupled transmission line models. The models are based on the electro-magnetic transient program (EMTP) power system transient analysis tools, extended to two dimensions. The new tool is targeted at acoustic wave propagation phenomena. The method, called TINA for transient insular nodal analysis, uses temporal interpolation and fractional latency to maintain synchronicity in heterogeneous media. The fractional latency method allows the model cells to operate at a local simulation time step which can be a non-integer ratio of the global simulation time step. This simplifies synchronicity and saves computation time and memory. Thévenin equivalents are used to interface the mesh cells and provide an abstraction of the cell content. Numerically, the method is of the transmission-line matrix (TLM) family. In the thesis, loss-less and distortion-less models are considered. The loss-less transmission line models are studied for their stability and numerical error, for which analytical expressions are derived based on the simulation parameters. A number of new relations were discovered and discussed. The TINA method is evaluated in 2D using acoustic experiments, and also a new method is proposed for obtaining impulse responses in time-domain simulation, based on a periodic, band-limited impulse signals.

View record

Power transformers are among the most critical of assets for electric utilities, in the financialimpact that their failure can bring. Asset Managers need to be able to determine theright time for replacement, refurbishment or relocation of these devices, with an increasingdegree of confidence, in order to minimize the total cost of operation over the equipments’life. This has brought a change from scheduled maintenance to condition based monitoring(CBM), where the state of the transformer is continuously monitored to evaluate its workingcondition.A key method of transformer CBM, which effectively detects mechanical damage to thestructure of the transformer windings, is Frequency Response Analysis (FRA). FRA relieson comparison of electrical admittance signatures to determine if the winding has becomedeformed. One of the major problems it still faces is the interpretation of differences inthe signatures. To date, experts are needed to analyse graphs, drawing from experience inorder to produce educated guesses as to what the differences in admittance functions denote.However, in the recent past, there has been some headway in programming computer basedsolutions for the problem of interpretation.The use of Artificial Neural Networks (ANNs) has perhaps been the most promising in thisrespect. ANNs perform in the same way that human experts do, drawing upon experience tomap a change in shape of a signature to a physical change in the winding system. However,one of the major drawbacks of these methods is the large training data-sets required for theneural network to learn.The work reported in this thesis seeks to address this problem by generating training datasetsfrom analytical models of the transformer. Due to the large number of simulations thatneed to be performed a customized solution method was developed to speed up computations.A combination of back propagation and radial basis function networks were then usedto classify the type, location and severity of winding movement. The results showed thatthe neural network approach was not only accurate but tolerant to high noise levels.

View record

Critical infrastructures are the lifelines of modern societies. The Communication and Information Technology Infrastructure (CITI) provides the basic mechanisms for sharing control and decision-making information among different critical infrastructures. Failures in CITI, either due to an accident or malicious action can propagate to other infrastructures and degrade or disrupt their functionality. Conversely, failures in other infrastructures can also propagate to CITI and hence disrupt the operation of many of the interconnected systems. For reliable and consistent operation of critical infrastructure networks, it is important to have tools and techniques to model and simulate CITI related interdependencies. This research is focusing on developing such methods and tools for CITI interdependency modelling and simulation. Our approach is based on system engineering techniques, where critical infrastructures are viewed as a system of systems. Interdependencies between different system components are captured using precise mathematical functions. As such, our approach goes beyond the limitations of agent-based modelling and simulation paradigms, where interdependencies are considered an emergent behavior. In this research, we have used predictive modelling techniques commonly used in power systems, data communication networks and information systems. The approach is based on results from real CITI interdependency related data. In our model, we used these data to identify the origin of different types of CITI failure and their impacts on critical infrastructures. Following that, we developed techniques to estimate interdependencies between CITI and other critical infrastructures. Finally, we developed techniques to simulate CITI interdependencies in a critical infrastructures simulator. The simulation results were validated against real-life failure cases. Our approach gives a comprehensive solution to CITI interdependency modelling and simulation problems and hence is an important step in the critical infrastructure related research. Even though our techniques are developed for CITI interdependency, they will be useful for other critical infrastructure networks as well.

View record

This thesis presents a method of detecting electrical islands in low voltage distributed generator networks by measuring negative sequence impedance differences between islanded and utility connections. Extensive testing was conducted on a commercial building and 25 kV distributed generator fed network by measuring naturally occurring and artificially injected negative sequence components. Similarly, this technique was tested using the IEEE 399-1990 bus test case using the EMTP software. The practical measurements have been matched to simulations where further system performance characteristics of detecting power system islands has been successfully demonstrated. Measured results indicate that unbalanced load conditions are naturally occurring and readily measurable while deliberately unbalanced loads can increase the accuracy of negative sequence impedance islanding detection. The typically low negative sequence impedance of induction motors was found to have only a small effect in low voltage busses, though large machines can effect the threshold settings. Careful placement of the island detector is required in these situations. The negative sequence impedance measurement method is an improvement on previous impedance measurement techniques for islanding detection due to its accuracy, and distinctly large threshold window which have challenged previous impedance based islanding detection techniques.

View record

The requirements of today's power systems are much different from the ones of the systems of the past. Among others, energy market deregulation, proliferation of independent power producers, unusual power transfers, increased complexity and sensitivity of the equipments demand from power systems operators and planners a thorough understanding of the dynamic behaviour of such systems in order to ensure a stable and reliable energy supply.In this context, on-line Dynamic Security Assessment (DSA) plays a fundamental role in helping operators to predict the security level of future operating conditions that the system may undergo. Amongst the tools that compound DSA is the Transient Stability Assessment (TSA) tools, which aim at determining the dynamic stability margins of present and future operating conditions.The systems employed in on-line TSA, however, are very much simplified versions of the actual systems, due to the time-consuming transient stability simulations that are still at the heart of TSA applications. Therefore, there is an increasing need for improved TSA software, which has the capability of simulating bigger and more complex systems in a shorter lapse of time.In order to achieve such a goal, a reformulation of the Multi-Area Thévenin Equivalents (MATE) algorithm is proposed. The intent of such an algorithm is parallelizing the solution of the large sparse linear systems associated with transient stability simulations and, therefore, speeding up the overall on-line TSA cycle. As part of the developed work, the matrix-based MATE algorithm was re-evaluated from an electric network standpoint, which yielded the network-based MATE presently introduced. In addition, a performance model of the proposed algorithm is developed, from which a theoretical speedup limit of p/2 was deduced, where p is the number of subsystems into which a system is torn apart. Applications of the network-based MATE algorithm onto solving actual power systems (about 2,000 and 15,000 buses) showed the attained speedup to closely follow the predictions made with the formulated performance model, even on a commodity cluster built out of inexpensive out-of-the-shelf computers.

View record

Electromagnetic Transients Program (EMTP) simulators are being widely used in powersystem dynamics studies. However, their capability in real time simulation of power systems iscompromised due to the small time step required resulting in slow simulation speeds.This thesis proposes a Shifted Frequency Analysis (SFA) theory to accelerate EMTPsolutions for simulation of power system operational dynamics. A main advantage of the SFA isthat it allows the use of large time steps in the EMTP solution environment to accuratelysimulate dynamic frequencies within a band centered around the fundamental frequency.The thesis presents a new synchronous machine model based on the SFA theory, whichuses dynamic phasor variables rather than instantaneous time domain variables. Apart from usingcomplex numbers, discrete-time SFA synchronous machine models have the same form as thestandard EMTP models. Dynamic phasors provide envelopes of the time domain waveforms andcan be accurately transformed back to instantaneous time values. When the frequency spectra ofthe signals are close to the fundamental power frequency, the SFA model allows the use of largetime steps without sacrificing accuracy. Speedups of more than fifty times over the traditionalEMTP synchronous machine model were obtained for a case of mechanical torque step changes.This thesis also extends the SFA method to model induction machines in the EMTP. Byanalyzing the relationship between rotor and stator physical variables, a phase-coordinate modelwith lower number of equations is first derived. Based on this, a SFA model is proposed as ageneral purpose model capable of simulating both fast transients and slow dynamics in inductionmachines. Case study results show that the SFA model is in excess of seventy times faster thanthe phase-coordinate EMTP model when simulating the slow dynamics.In order to realize the advantage of SFA models in the context of the simulation of thecomplete electrical network, a dynamic-phasor-based EMTP simulation tool has been developed.

View record