Search

This article discusses how techniques from reinforcement learning (RL) can be exploited to transition to a model-free and scalable wide-area oscillation damping control of power grids. We present two control architectures with distinct features. Performing full-dimensional RL control designs for any practical grid would require an unacceptably long learning time and result in a dense communication architecture. Our designs avoid the curse of dimensionality by employing ideas from model reduction. The first design exploits time-scale separation in the generator electro-mechanical dynamics arising from coherent clustering, and learns a controller using both electro-mechanical and non-electro-mechanical states while compensating for the error in incorporating the latter through the RL loop. The second design presents an output-feedback approach enabled by a neuro-adaptive observer using measurements of only the generator frequencies. The controller exhibits an adaptive behavior that updates the control gains whenever there is a notable change in the loads. Theoretical guarantees for closed-loop stability and performance are provided for both designs. Numerical simulations are shown for the IEEE 68-bus power system model. [ABSTRACT FROM AUTHOR]

In this tutorial we present the state-of-the-art as well as new methods for simulating various planning, operation, stability, reliability, and economic models of electric power systems. The discussion is driven by both first-principle models and empirical models. First-principle models result from the fundamental physics and engineering principles that govern the behavior of various components of a grid. Empirical models, on the other hand, are models that result from statistics and data analysis. We overview a wide spectrum of applications starting from planning models with a time-scale of simulation in years to real-time models where the time-scale can be in the order of milliseconds. We present a list of simulation software popularly used by the power engineering research community across the world. The increasingly important roles of power electronics, communication and computing, model aggregation, hybrid simulation, faster-than-real-time simulation, and co-simulation in emulating the daily operation of a grid are enumerated. The importance of research testbeds for testing, verification and validation of complex grid models at various temporal and spatial scales is also highlighted. The overall goal is to provide a vision on how simulations and their supporting implementation methods can help us in understanding the evolving behavior of tomorrow’s power networks as a truly intelligent cyber-physical system. [ABSTRACT FROM PUBLISHER]

In a recent paper, we proposed a system identification method for constructing reduced-order models for the electro-mechanical dynamics of large power systems, divided into multiple coherent clusters, using Synchrophasors. Every cluster in the actual model was represented as an aggregate generator in the reduced-order model. An aggregate network graph connected one aggregate generator to another. In this paper, we extend this identification approach to differential-algebraic (DAE) models. First, every cluster is associated with a unique terminal bus, referred to as the pilot bus, that couples its internal network to the rest of the system. The proposed algorithm uses Synchrophasor measurements from the pilot buses to identify the dynamic model of the aggregate generator for each cluster using nonlinear least squares while retaining the identity of all the pilot buses. The resulting reduced-order model is in the form of a nonlinear electric circuit described by aggregate differential and algebraic equations. We illustrate our results using two case studies, one for the IEEE 9-bus power system and another for the IEEE 39-bus power system. We also discuss how these reduced-order DAE models may be useful for designing shunt controllers at the pilot buses by using Synchrophasor feedback. [ABSTRACT FROM AUTHOR]

We develop a graph-theoretic algorithm for localizing the physical manifestation of attacks or disturbances in large power system networks using real-time synchrophasor measurements. We assume the attack enters through the electro-mechanical swing dynamics of the synchronous generators in the grid as an unknown additive disturbance. Considering the grid to be divided into coherent areas, we pose the problem as to localize which area the attack may have entered using relevant information extracted from the phasor measurement data. Our approach to solve this problem consists of three main steps. We first run a phasor-based model reduction algorithm by which a dynamic equivalent of the clustered network can be identified in real-time. Second, in parallel, we run a system identification in each area to identify a transfer matrix model for the full-order power system. Thereafter, we exploit the underlying graph-theoretic properties of the identified reduced-order topology, create a set of localization keys, and compare these keys with a selected set of transfer function residues. We validate our results using a detailed case study of the two-area Kundur model and the IEEE 39-bus power system. [ABSTRACT FROM PUBLISHER]

In this paper, we present an optimization algorithm that selects the optimal sets of points for placing phasor measurement units (PMUs) on the transmission lines of a multimachine power system for the purpose of identifying the best model fit for its wide-area swing dynamics. Alternatively, the method can also be viewed as a way to select the optimal set of points at which phasor values should be computed using measurements available from PMUs such that these computed values, also referred to here as pseudo-measurements, can generate the best estimate of the swing model, especially when the measurements are noisy. We pose the identification problem as an equivalent parameter estimation problem for the admittance of each tie-line and the inertia of each machine using phasor measurements of voltage magnitude, phase angle and frequency, corrupted with high-frequency measurement noise. We then formulate the Cramer–Rao bounds (CRBs) for the estimates of these unknown parameters, and show that the bounds are functions of the PMU locations and of the contribution of each measurement variable in the combined output. We finally state the condition for finding the optimal PMU location that guarantees the tightest CRB, and, therefore, the most accurate swing model for the system. [ABSTRACT FROM PUBLISHER]

This brief presents an analytical method to study the effects of wind farm location on the interarea oscillation spectrum of large radial power systems along with a means to shape this spectrum. We consider a continuum representation of the electromechanical swing equations for the power system with the wind power injection modeled as a spatial point source forcing applied at a certain electrical distance from one end of the transfer path. The resulting forced hyperbolic wave equation for the rotor phase angle is used to derive an analytical expression for the spectrum of the power flow as a function of the wind power injection location. A simulation study shows that judicious siting of the wind farm can provide effective damping to a particular set of modes. Finally, we show how the analysis framework can be extended to provide a means of shaping the interarea oscillation spectrum for a grid with a wind farm at an arbitrary location using a simple optimal control design for the wind power output. [ABSTRACT FROM AUTHOR]

In this paper we present a FACTS (Flexible AC Transmission Systems)-based control design for electromechanical oscillation damping in large power systems, facilitated by aggregate models that can be constructed using Synchronized phasor measurements. Our approach consists of three steps, namely—1. Model Reduction, where Synchrophasors are used to identify second-order models of the oscillation clusters of the power system retaining the inter-ties on which FACTS devices such as Thyristor Controlled Series Compensators (TCSC) are installed, 2. Aggregate Control, where feedback controllers are designed to achieve a desired closed-loop transient response between every pair of clusters, and finally 3. Control Inversion, where the aggregate control design is distributed and tuned to actual TCSC controllers in the full-order model until its inter-area responses match the respective inter-machine responses of the reduced-order system. It is shown that the inversion problem can be posed equivalently as decomposing the swing dynamics into fast and slow states, and designing the controllers such that the slow dynamics can optimally track a desired closed-loop signal designed for the aggregate model. Application of the approach to two-area power systems is demonstrated through topological examples inspired by the US west coast grid. [ABSTRACT FROM PUBLISHER]

Wide-area analysis and control of large-scale electric power systems are highly dependent on the idea of aggregation. For example, one often hears power system operators mentioning how northern Washington oscillates against southern California in response to various disturbance events. The main question here is whether we can analytically construct dynamic electromechanical models for these conceptual, aggregated generators representing Washington and California, which in reality are some hypothetical combinations of thousands of actual generators. In this paper we address this problem, and present a concise overview of several new results on how to construct simplified interarea models of large power networks by using dynamic measurements available from phasor measurement units (PMUs) installed at specific points on the transmission line. Our examples of study are motivated by widely encountered power transfer paths in the Western Electricity Coordinating Council (WECC), namely, a two-area radial system representing the WA-MT flow, and a star-connected three-area system resembling the Pacific AC Intertie. [ABSTRACT FROM PUBLISHER]