CNLS Smart Grid Seminar Series
Sponsored by CNLS, IS&T, Energy Institutes at LANL
& LDRD DR on ``Optimization and Control Theory for Smart Grids"
CNLS conference room, Tues, 10:30-12:00
| Ensemble control involves the manipulation of an uncountably in nite collection of structurally identical or similar dynamical systems, which are indexed by a parameter set, by applying a common control without using feedback. This subject is motivated by compelling problems in quantum control, sensorless robotic manipulation, and neural engineering, which involve ensembles of linear, bilinear, or nonlinear oscillating systems, for which analytical control laws are infeasible or absent. My focus is on constructive control design methods for practical ensemble control problems. The fi rst result is an efficient numerical method based on the singular value decomposition (SVD) for the synthesis of minimum norm ensemble controls for time-varying linear systems. This method is extended to iterative techniques to accommodate bounds on the control amplitude, and to synthesize ensemble controls for unitary bilinear systems. Example ensemble systems include harmonic oscillators, quantum transport, and quantum spin transfers on the special orthogonal group SO(n), in particular the Bloch system on SO(3). Another result involves the control of ensembles of nonlinear oscillators, which occur in neuroscience and electrochemistry. The ability to optimally manipulate such systems provides insight into treatments for disorders such as Parkinson's disease and epilepsy. A key phenomenon is entrainment, which refers to the dynamic synchronization of an oscillating system to a periodic input. Phase coordinate transformation, formal averaging, and the calculus of variations are used to derive minimum energy and minimum mean time controls that entrain ensembles of non-interacting oscillators to a harmonic or subharmonic target frequency, and establish desired dynamical structures. |
| During the last years a fundamental transformation of the electric power and integrated energy systems has been initiated in Europe and other industrialized countries. These new developments will drastically change the structure of these systems and the way they are operated. One can identify two main driving forces in this process. First, a massive introduction of distributed renewable power sources, i.e. mostly wind power and photo voltaics (PV), requires new system solutions. Since these sources are fluctuating and uncertain new methods for planning, managing, and operating the system must be developed and introduced. Second, information and communication technologies (ICT) offer new possibilities with regard to system control in general and management of distributed power sources and demand side response in particular. This presentation will give an overview of the current developments in this field. The work at ETHZ concerning modeling of future energy systems will be presented. In particular the energy hub and power node will be described and their use in system analysis exemplified. Simulations from real system will be presented. The role of storage devices and demand side response will be elaborated. |
| Power systems are undergoing an unprecedented period of change driven by increased penetration of renewable generation, electric vehicles, and increased take-up of Demand Side Response facilitated by smart metering. All those changes mean that the traditional mode of power system operation and control based on centralised control, deterministic (N-1)-based dispatch and generation following demand will have to change. A new paradigm is needed to facilitate distributed and stochastic control. This change of paradigm will require a significant interdisciplinary effort combining the skills of engineers, mathematicians, economists, social scientists and politicians. The talk will give examples of some of the research projects that address those challenges. |
| Future smart grids will likely support bi-directional flow of electricity and include power production from multiple, disparate, and uncontrollable sources due to a high penetration of distributed renewable energy resources. Some of the more challenging problems for the future grid include maximizing the use and efficiency of renewable resources, and realizing optimal demand and power production responses that can complement renewable intermittency. Integration of renewables together with energy storage systems has been motivated by the increasing attention to feature renewable energies from not only solar and wind power but also the excess generation from many customers. Effective use of renewable resources using battery systems can be realized by balanced distribution of such distributed energy resources (DERs) with complementary demand and dispatchable generation responses. The spatial distribution, intermittency, and uncontrollability of most renewable resources, however, make stable and reliable electricity transmission and distribution difficult especially with high renewable market penetration in large-scale complex power networks. In order to use energy storage systems effectively to optimize DERs as well as realize a reliable and sustainable future grid, we present an autonomous distributed management model that can realize optimum power flow control together with demand and power response, which especially integrates Kirchhoffs core theory and autonomous agent systems. |
| In recent years, UAVs been have extensively used as relays in disaster management applications. By the dynamic nature of UAVs as they traverse rugged terrains, the problem of determining the interconnections among UAVs is important from the view point of power consumption, maximum number of communication links and robust connectivity. We pose this network synthesis problem as a Mixed Integer Semi-Definite Program (MISDP) with the smallest non-zero eigenvalue of the weighted networkss Laplacian matrix as the measure for robustness. Solving this MISDP is a difficult optimization problem because of its non-linear objective coupled with the possibility that the number of feasible solutions combinatorially explode with the size of the network. In this presentation, we discuss novel algorithms based on cutting plane methods to obtain optimal solutions and upper bounds for problems of moderate sizes. Also, based on the spectrum of connected networks, we develop efficient neighbourhood search heuristics which can be applied for large UAV networks. |
| The equilibrium operation of power grid requires that the generation meets the demand at each time instant and any deviation raises critical stability concerns. However with ever increasing load, more so with the imminent release of plug-in electric vehicles en masse, the grids are under greater pressure. On the other hand, the growing penetration of renewable energy sources provides an excellent opportunity to meet the increased electricity demand, but the challenge remains to mitigate the uncertainties associated with renewable generation. The challenge here is to ensure seamless integration of newer forms of generation and load, while maintaining a stable and satisfactory grid-level performance. Specifically in this talk, I will be covering how we propose to model the aggregate dynamics of a large group of flexible (or time deferrable) electric loads, such as plug-in electric vehicle chargers, thermostat-controlled heating/cooling loads, etc. using a hysteresis-based approach, and control their aggregate electricity demand to mitigate fluctuations in renewable generation. It will be shown that, often this population dynamics exhibits interesting nonlinear behavior, such as period adding cascade, and thus need to be understood well to ensure safe electrical grid operations. Finally I would briefly discuss some of the more recent techniques that concern with analyzing the power systems stability from a Lyapunov stability perspective, which is often very complex because of the systems complexity and high dimensionality. However a recent advancement shows promising results by using sum-of-squares techniques to compute the systems Lyapunov function. |
| We consider the problem of synthesizing optimal linear feedback policies subject to arbitrary convex constraints on the feedback matrix. This is known to be a hard problem in the usual formulations (H2; H1; LQR) and previous works have focused on characterizing classes of structural constraints that allow ef?cient solution through convex optimization or dynamic programming techniques. In this paper, we propose a new control objective and show that this formulation makes the problem of computing optimal linear feedback matrices convex under arbitrary convex constraints on the feedback matrix. This allows us to solve problems in decentralized control (sparsity in the feedback matrices), control with delays and variable impedance control. Although the control objective is nonstandard, we present theoretical and empirical evidence that it agrees well with standard notions of control. We also present an extension to nonlinear control af?ne systems. We present numerical experiments validating our approach. |
| Due to the potential for finding globally optimal solutions, significant research interest has focused on the application of semidefinite optimization techniques to problems in the field of electric power systems. This seminar discusses a semidefinite relaxation of the non-convex AC optimal power flow (OPF) problem, which seeks to minimize the operating cost of an electric power system subject to both engineering inequality and network equality constraints. The convex semidefinite relaxation is capable of finding globally optimal solutions to many OPF problems. By exploiting power system sparsity, semidefinite relaxations of practically sized OPF problems are computationally tractable. The semidefinite relaxation is tight for many but not all OPF problems. For practical problems where the semidefinite relaxation is not tight, results show small active and reactive power mismatches at the majority of load buses while only small subsets of the network exhibit significant mismatch. This suggests that the relevant non-convexities in these problems are isolated in small subsets of the network. Examination of the feasible spaces for small test cases illustrates such non-convexities and explains the semidefinite relaxations lack of tightness. Finally, preliminary results from the application of higher-order moment semidefinite relaxations show promise in obtaining globally optimal solutions to these small test cases. |
| There is an emerging consensus that the nation's electricity grid is vulnerable to cyber attacks. This vulnerability arises from an increasing reliance on transmitting remote measurement data over legacy data networks to system operators who make critical decisions based on available data. Data integrity attacks are a class of cyber attacks that involve a compromise of information that is processed by the grid operator. These attacks have consequences only when the system operator responds to compromised data, for example, by re-dispatching generation under normal or contingency protocols. These consequences include (a) financial losses from sub-optimal economic dispatch to service loads, (b) robustness/resiliency losses from placing the grid at operating points that are at greater risk from contingencies, and (c) systemic losses resulting from cascading failures induced by poor operational choices. In this talk we compute the worst case economic consequence of an unobservable data integrity attacks. This serves as an effective metric to assess the importance of various attacks. |
| Energy Systems Integration is a methodology for deliberate and objective energy system planning, operations, and optimization across multiple scales, domains, and time resolutions. Dr. Kroposki will discuss the vision for the future energy system to develop, demonstrate, and operate highly integrated, flexible, scalable, and efficient systems that provide integration of clean energy sources while maintaining reliability and resiliency at an affordable cost. Dr. Kroposki will also discuss NRELs recently opened Energy Systems Integration Facility (ESIF). At 182,500ft2, the ESIF is the largest R&D facility on NRELs campus and focuses on research and development of clean energy technologies such as variable renewable generation, smart load controls, and electric vehicles that are being deployed in the electric power system at an increasing rate. ESIF research seeks to connect the simulation environments with demonstration of technology at scale through hardware-in-the-loop testing. |
| Nowadays, the world is facing several major challenges that include air and water pollution, global warming and the rising market prices of the primary energy resources, among others. Natural gas, as an energy source, offers several advantages in comparison with other non-renewable energy sources to overcome these problems. For example, natural gas is a cleaner fossil fuel than oil or coal, i.e., it emits a lower percentage of carbon dioxide than gasoline, diesel or coal. Natural gas is also more economically attractive than gasoline despite that its listing on the financial sector has been increasing in recent years (which represents better profits for the industrial sector). Since natural gas has become a good candidate for being one of the preferential supplies of primary energy, the natural gas industry has had to quickly expand its transmission networks in order to satisfy the increasing demand of the gas consumption. Hence, this presentation aims at integrating mathematical models and solution approaches for tackling various optimization problems in natural gas transport via pipeline systems. Mainly, three challenging problems and their underlying optimization methods are addressed: (1) The fuel cost minimization problem -formulated as a non-linear programming (NLP) model- for which three different solution methodologies are proposed, namely, (a) a heuristic method that includes a non-sequential dynamic programming technique, (b) a tree decomposition technique and a dynamic programming algorithm, which is proposed to overcome dense network instances, and (c) an adaptive discretization (multi-local search) heuristic to enhance the application of the dynamic programming. (2) Natural gas transport problems with variable specific gravity and compressibility factor. Here, an enhanced mathematical model is proposed to account for more accurate estimates in maximum flows on steady-state transmission network systems. This problem arises since traditional approaches in steady-state flow problems assume the gas specific gravity and compressibility factor as universal constants, thus leading to misleading results. Due to the non-convexity of the suggested model, a heuristic algorithm based on an iterative scheme is proposed in which a simpler NLP model is solved. (3) The line-packing problem. Here, a mathematical model is proposed to optimize the short-term storage and transport of natural gas in pipelines for a given planning horizon. The proposed model adopts all characteristics of a mixed-integer non-linear programming (MINLP) model. A thorough computational evaluation based on a global optimizer is conducted to assess the computability of the model. Empirical evidence over a wide set of problem instances illustrate the usefulness and positive impact of the proposed strategies resulting in cconsiderably high-quality solutions when compared to existing approaches and commercial methods. |
| Energy microgrids fully integrated with buildings and with the smart-grid are a promising concept for accelerating the introduction of distributed energy generation. When fully implemented, this concept will provide a large number of benefits ranging from a wider use of renewable resources to improved energy efficiency, power quality and reliability. In order to bring this concept to market, there are a large number of technologies and systems integration concepts that need to be mature alongside with the development of a strong business case for the involved stakeholders. These technologies include microgrid energy and power management systems, alongside with many other enabling technologies such as smart meters, power electronics, communications between stakeholders and microgrid components, implementation of cyber-security, etc. The energy management system (or supervisory system) is responsible for decisions relative to supply and demand energy flows and set-points based on operating costs, customer preferences, utility requests and operational constraints, and communicates these decisions by dispatching set-points to the local controllers. The power management system requires a much higher bandwidth, and needs to provide system stability, coordination between multiple microgrid components, and synchronous connection and disconnection with the grid; alongside with the capability to provide other power services such as power factor and power quality correction. The seminar to be presented at Los Alamos National Lab will introduce some of the power and energy management systems that United Technologies Research Center (UTRC) has been developing and demonstrating, as part of its integrated building solutions portfolio. UTRC has been working with concepts based on model predictive controls and stochastic programming formulation (SPF) to address the uncertainty of load and weather profiles and the dynamic nature of energy storage and renewable resources. Besides, power algorithms for seamless transition with the external-grid, providing multiple ancillary and power services have been demonstrated in real buildings and will be presented. |
| Modern society critically depends on the services electric power provides. Power systems rely on a network of power lines and transformers to deliver power from the sources of power (generators) to the consumers (loads). However, when power lines fail (for example through lightning or natural disasters) the network is often NOT able to fulfill all of the demand for power. To mitigate these failures, increasingly, new devices such as FACTS devices have been deployed on power systems. A FACTS device allows power grid operators to adjust the impedance parameters of power lines, thereby redistributing flow in the network and potentially increasing the amount of power that is supplied. Here we develop new approaches for determining the optimal parameter settings for supplying the maximal amount of power. |
| Power system operators make real time control decisions based on the real time state estimate. This state includes not only continuous variables (complex voltage at each bus) but also discrete topological state variables. Previous work has developed uncertainty quantification methods that provide bounds on the continuous state variables which hold at specific confidence levels. I will present an extension of this work that allows for estimation of discrete topological variables. The objective of this project is to provide a topological model bank that the true topology must lie in. Furthermore, we can assure that the true topology falls within this model bank at a specific certainty. |
| Demand response (DR) is increasing being viewed as an alternative method for balancing generation and load on the power system. To this end, HVAC loads, which incorporate a significant amount of thermal storage, could potentially provide this service with a minimal investment in additional infrastructure. This research looks at a large commercial building that has been equipped for DR and tests the system dynamics under various control inputs. The results suggest that building energy use and efficiency are significantly impacted by the type of control signal applied, implying that care must be taken to reduce transient losses in a DR application. |
| The electric power grid is undergoing rapid changes driven by demand growth, rising energy costs, concerns about energy security and the desire to integrate more renewable energy sources. In order to facilitate these changes a greater understanding of how they will affect both the stability and performance of the power system is required. For example, the addition of inherently intermittent renewable energy sources such as solar and wind power will affect the power balance on the grid. The nature of these resources also has the potential to make the power system more distributed through the addition of numerous small wind and solar plants. This talk illustrates the use of control and optimization based methods to provide insight into a few example problems related to the design, operation and management of the envisioned new power system. First, we discuss the use of storage to provide greater system flexibility and to mitigate the inherent variability of renewable sources. We then extend this idea to investigate the factors that drive optimal storage sizing and siting in a transmission network. The second part of the talk will briefly introduce two complementary problems. The first examines how increasing amounts of distributed generation will affect power system efficiency. In particular, we evaluate the losses associated with synchronizing a power system after a transient disturbance or in maintaining synchrony in the face of ongoing disturbances and their relationship to the network properties. Finally, we address the question of how wind farm placement affects system damping and offers control strategies to drive the frequency response of the integrated system to a desired shape. The array of problems discussed represent results and analysis for a small subset of stability and performance issues related to grid efficiency and are meant to demonstrate the fact that achieving the full potential of smart and clean power systems is a multifaceted problem that will require a combination of strategies. |
| In this tutorial I will summarize recent developments on the convex relaxations of optimal power flow (OPF) problems. OPF is a fundamental problem that underlies many power system operations. It can be formulated as a nonconvex quadratically constrained quadratic program (QCQP). Recently several convex relaxations have been developed based on semidefinite programming, chordal extension, and second-order cone programming in both bus injection model and branch flow model. I will explain the relations among these relaxations, and the various sufficient conditions in the literature that guarantee the exactness of these relaxations. |
| Continual growth in the size and peak power consumption of high performance computing (HPC) platforms is increasing the stress on local power distribution systems. In particular, the large, fast and random transitions in HPC power consumption create large and uncertain voltage swings and power loss, unless proper control is designed and performed. We consider three types of control devices, i.e., fixed capacitor, switchable capacitor and FACTS device, installed on the HPC load side of the power distribution system. Though these devices have different control logics and response characteristics, they can jointly provide the reactive power compensation required by HPC to regulate the voltage and decrease power loss (improve energy efficiency). We formulate the minimum power loss objective and the voltage safe bound as a chance-constrained optimization problem for HPC power system control. Moreover, supposing that the optimal control is always performed and considering different prices of control devices, we formulate an optimal control devices sizing problem. Based on the statistics of HPC power transitions observed from LANL's Ceilo platform, we find the structures of both the control and sizing problems which make them tractable to solve. This is a joint work with Misha Chertkov, Scott Backhaus and Steven Low. |
| Natural gas is used to heat homes and to power gas-turbines in power plants which produce electricity. Sources of natural gas are often separated by great distances from the loads. As a result, there are major gas pipelines that run across states and across countries. Laws of physics that govern the steady-state flow through these pipelines dictate that the square flow is proportional to the difference in square pressure between the ends of a pipe and inversely to the length of the pipe. We consider networks with tree structures, which closely resemble the structure of major interstate pipelines in the US. Given a fixed input flow, the remaining flows on the tree are uniquely determined based on the loads. Since it is not uncommon for pipeline lengths to exceed 1,000 miles, to prevent pressure from dropping too much it is necessary to install compressor stations along the pipe which locally boost the pressure, making it feasible to transport the gas over such long distances. However, there is an operational cost associated with running the compressors that depends on their compression ratios: the ratio of outlet to inlet pressure at the compressor. Different configurations of compressor ratios might lead to feasible pressures that support the flows, but some are more expensive than others. The goal is to find an optimal configuration that minimizes the total cost of running the compressors while maintaining feasible pressures. We propose two ways to solve this optimization problem efficiently. The first method is based on reformulation of the problem as a geometric program, and the second is based on a well-known dynamic programming approach. We apply both these methods to the Belgium gas network and to the US Transco pipeline, which runs from the Gulf of Mexico up to Pennsylvania, and discuss the results. |
| The contribution of renewable resources to the energy portfolio across the world has been steadily increasing over the past few years. Several studies predict the continuation of this trend in the future leading to large scale integration of renewable resources into energy networks. A principal challenge associated with this is the intermittency and non-dispatchability of the renewable sources. This necessitates the need to incorporate faster reserves, storage devices and similar services operating alongside the slow ramping conventional generators in the energy network. To maintain the robustness of such a network, there are proposals to require hourly planning for some resources, and sub-hourly planning for others: an hourly scale may be used for conventional generator production levels and a sub-hourly scale for renewable generator levels and/or storage and transmission network utilization. The talk will present a multiple time scale stochastic programming formulation of the economic dispatch problem and algorithmic frameworks to tackle it. The first approach highlights the difference between hourly and sub-hourly planning of economic dispatch and uses the two-stage Stochastic Decomposition(SD) algorithm. The second framework combines three principal components: optimization, dynamic control and simulation. The conventional generator decisions are obtained iteratively by solving a regularized linear problem in the first stage of SD. For these first stage decisions, a policy for recommending the dispatch decisions is identified using an Approximate Dynamic Programming based controller. A vector auto-regression based simulator is used to provide the sub-hourly wind generation scenarios. The performance of these algorithms was tested on the IEEE model networks and the Illinois network. The insights gained regarding the benefits of sub-hourly planning and role of operating reserves/storage in energy network with high renewable penetration will be presented. |
| Inter-area oscillations in bulk power systems are typically poorly controllable by means of local decentralized control. Recent research efforts have been aimed at developing wide-area control strategies that involve communication of remote signals. In conventional wide-area control, the control structure is fixed a priori typically based on modal criteria. In contrast, here we employ the recently-introduced paradigm of sparsity- promoting optimal control to simultaneously identify the optimal control structure and optimize the closed-loop performance. To induce a sparse control architecture, we regularize the standard quadratic performance index with an L1-penalty on the feedback matrix. The quadratic objective functions are inspired by the classic slow coherency theory and are aimed at imitating homogeneous networks without inter-area oscillations. We briefly review the slow coherency theory and illustrate it with different examples. Next, we use the New England power grid model to demonstrate that the proposed combination of the sparsity-promoting control design with the slow coherency objectives performs almost as well as the optimal centralized control while only making use of a single wide-area communication link. In addition to this nominal performance, we also demonstrate that our control strategy yields favorable robustness margins and that it can be used to identify a sparse control architecture for control design via alternative means. |
| In this presentation we review modeling of modern power markets, its applications, methods, software tools and challenges faced by typical users of these tools. Special emphasis is placed on the ability of modeling tools to replicate the multi-cycle decision logic underlying operations of real power markets spanning over multiple time scales ranging from minutes to days and weeks. We will review the implementation of such decision logic in the Power System Optimizer (PSO). Next, we discuss the opportunities provided by commercial cloud computing in modeling power markets and overview the architecture of pCloud, a power market simulation environment utilizing the PSO engine and implemented on the Windows Azure and Amazon commercial clouds. |
| We formulate an optimal load control (OLC) problem in power networks where the objective is to minimize the aggregate cost of tracking an operating point subject to power balance over the network. We prove that the swing dynamics and the branch power ?ows, coupled with frequency-based load control, serve as a distributed primal-dual algorithm to solve OLC. Even though the system has multiple equilibrium points, we prove that it nonetheless converges to an optimal point. This result implies that the local frequency deviations at each bus convey exactly the right information about the global power imbalance for the loads to make individual decisions that turn out to be globally optimal. It allows a completely decentralized solution without explicit communication among the buses. Simulations show that the proposed OLC mechanism can resynchronize bus frequencies with signi?cantly improved transient performance. |
| With the increasing penetration of renewable energies, the power system is being stressed by the great uncertainty from the supply side. On the other hand, with the popularity of electric vehicles and the expansion of data centers, the power system will witness a significant increase of demand in the next decade. Either of these two stresses alone can make the conventional power system collapsed easily. Together, they imply smart grid is a crucial task in the next decade. A naive approach to solve the problems is to directly ask the demand to follow the supply in a centralized control fashion. However, this may not work in practice since each entity in the system pursues its own interest. This motivates us to study the deregulated electricity market for the smart grid from a network economic approach. To ensure a successful market, we investigate two key challenges: efficiency and fairness. To understand the first challenge, we study the ancillary service market. One important challenge with wind DG units is to provide low-cost and fast-responding reactive power compensation of the wind turbine's inductive load to ensure a stable voltage profile in the system. Though STATCOMs have fast enough response time, they are usually expensive and may not be a feasible solution for large-scale deployment of wind DG units. We look at an alternative approach to compensate reactive power of wind DG units: to utilize the the inverter circuits in the charger of PEVs. We consider a scenario where a wind DG unit is co-located with a PEV charging station, and we use game theoretic model to ensure adequate incentives to the PEV owners to actively participate in the market. Our incentive design can achieve the same optimal performance as the centralized control does. To tackle the second challenge - the fairness issue, our goal is to identify and to assess the market power in the deregulated electricity market. This is challenging because congestion fragments the transmission system into smaller zones, behind bottleneck interconnects, and the markets within these zones may be highly concentrated even when the whole transmission system seems competitive. We introduce a novel functional approach to measuring long term market power that unifies a variety of popular market power indices. Our functional approach naturally defines a family of superadditive market power measures and can serve as the guidance for the evolution of the power system. |