Yannan Sun

Calibrating multi-machine power system parameters with the extended Kalman filter

Large-scale renewable resources and novel smart-grid technologies continue to increase the complexity of power systems. As power systems continue to become more complex, accurate modeling for planning and operation becomes a necessity. Inaccurate system models would result in an unreliable assessment of system security conditions and could cause large-scale blackouts. This motivates the need for model parameter calibration, since some or all of the model parameters could either be unknown or inaccurate. In this paper, the extended Kalman filter is used to calibrate the parameters of a multi-machine power system in the presence of faults. The calibration performance is tested under varying fault locations, parameter errors, and measurement noise giving an insight into how many generators and which generators could be difficult to calibrate.

Reduced-order modeling of aggregated thermostatic loads with demand response

Demand Response is playing an increasingly important role in smart grid control strategies. Modeling the dynamical behavior of a large population of appliances under demand response is especially important to evaluate the effectiveness of various demand response programs. In this paper an aggregate model is proposed for a class of second-order Thermostatically Controlled Loads (TCLs). The model efficiently includes statistical information of the population, systematically deals with heterogeneity, and accounts for a second-order effect necessary to accurately capture the transient dynamics in the collective response. A good performance of the model however requires a high state dimension which dramatically complicates its formal analysis and controller design. To address this issue, a model reduction approach is developed for the proposed aggre-gate model, which can significantly reduce its complexity with small performance loss. The original and the reduced-order aggregate models are validated against simulations of thousands of detailed building models using GridLAB-D (an open-source distribution simulation software). The results indicate that the reduced-order model can accurately reproduce the steady-state and transient dynamics generated by GridLAB-D simulations with a much reduced complexity.

Distributed dynamic state estimation with extended Kalman filter

Increasing complexity associated with large-scale renewable resources and novel smart-grid technologies necessitates real-time monitoring and control. Our previous work applied the extended Kalman filter (EKF) with the use of phasor measurement data (PMU) for dynamic state estimation. However, high computation complexity creates significant challenges for real-time applications. In this paper, the problem of distributed dynamic state estimation is investigated. One domain decomposition method is proposed to utilize decentralized computing resources. The performance of distributed dynamic state estimation is tested on a 16-machine, 68-bus test system.

PMU placement for dynamic state tracking of power systems

The complexity of power systems continue to increase as load demands grow and new energy technologies emerge. Efficient methodologies and instrumentation are needed for real-time monitoring and control of power systems. Accurately tracking the state variables (rotor angle and speed) is necessary for monitoring system stability conditions and assessing the risks of large-scale system collapse. Previous work proposed an extended Kalman filter (EKF) method, which makes use of data from phasor measurement units (PMU) and corrects the estimation predicted by the system model, for real-time tracking of system dynamics. This paper will explore how the number and locations of PMUs installed in the system should be determined to ensure satisfactory performance of the EKF-based tracking. Finding the optimal PMU placement, i.e., attaining whole system observability with the fewest PMUs, is very difficult to solve. In this paper, a novel search algorithm is presented for determining PMU placement (location and quantity). The algorithm determines a placement that gives small tracking error in polynomial time, while the optimal placement would be determined in exponential time. A modified, scalable algorithm is also presented. Observability of grid dynamics is considered in the sense that all the state variables can be tracked dynamically. Furthermore, observability in the presence of faults is considered. Simulation results for a 16-machine system and a 50-machine system are provided.

Optimization of Customer Subscription Rates to Electric Utility Tariffs

In 2006 a test of residential electricity pricing using multiple tariffs was performed using a mixture of fixed-prices, time of-use prices and 5-minute real-time prices. Each tariff has advantages and disadvantages for both consumers and utilities. This paper examines a numerical optimization method to efficiently identify the optimal combination of customers to enroll under the available tariffs such that a utility can minimize the cost associated with both serving energy and the cost associated with the uncertainty of peak loads.

A distributed cooperative power allocation method for campus buildings

We propose a coordination algorithm for cooperative power allocation among a collection of commercial buildings within a campus. We describe a typical commercial building Heating, Ventilation, and Air Conditioning (HVAC) system, and utilize Model Predictive Control (MPC) to characterize its power flexibility. The power allocation problem is formulated as a cooperative game using the Nash Bargaining Solution (NBS) concept, in which buildings cooperatively maximize the product of their utilities subject to their local flexibility constraints and a total power limit set by the campus coordinator. To solve the optimal allocation problem, a distributed protocol is designed using dual decomposition of the Nash bargaining problem. Numerical simulations are performed to demonstrate the efficacy of our proposed allocation method.

Improved controller design of Grid Friendly™ Appliances for primary frequency response

The Grid Friendly™ Appliance (GFA) controller, developed at Pacific Northwest National Laboratory, was originally designed to autonomously switch off appliances by detecting under-frequency events. In this paper, the feasibility of using the GFA controller to provide primary frequency response is investigated. In particular, the impacts of an important design parameter, i.e., curtailing frequency threshold, on the primary frequency response are carefully analyzed for different situations. In the normal situation, the current method of selecting curtailing frequency thresholds for GFAs is found to be insufficient to guarantee the desired performance especially when the frequency deviation is shallow. In the extreme situations, the power reduction of online GFAs could be so excessive that it can even impact the system frequency negatively. As the first step towards the efforts to make GFAs suitable for providing primary frequency response, the existing controller design is improved by modifying the strategy of selecting curtailing frequency thresholds to ensure the effectiveness of GFAs in the normal situation.

Auction design for power markets based on standardized contracts for energy and reserves

The wholesale power markets have undergone substantial change over the past few years to deal with increasing penetrations of variable energy resources. The market design and processes have become more complex because of introduction of new products and services to deal with the variability and intermittency of renewable resources. Additionally, the increasing adoption of distributed energy resources presents new challenges for the power grid operations, requiring additional flexibility to ensure stable, reliable, and resilient grid operations. The current market designs, however, do not explicitly value flexibility from resources. This paper presents a novel unit commitment/economic dispatch problem formulation that explicitly values the flexibility offered by resources. The market design is based on the concept of two-part pricing where resources are first paid to make their capacities available, including the flexibility, in the day-ahead markets and subsequently paid performance payments based on real-time dispatch.

Improving Distribution Resiliency with Microgrids and State and Parameter Estimation

............................................................................................................................................ iii Acknowledgments ............................................................................................................................... v Acronyms and Abbreviations ...........................................................................................................vii 1.0 Introduction ................................................................................................................................ 1 1.1 Traditional Mitigation ........................................................................................................ 1 1.2 Distribution Resiliency ....................................................................................................... 1 2.0 Microgrids as a Resiliency Resource .......................................................................................... 3 2.1 Previous Work .................................................................................................................... 3 2.2 Asset Feasibility ................................................................................................................. 3 2.2.1 Dynamic considerations .......................................................................................... 4 2.2.2 Distribution line and transformer in-rush ................................................................ 7 2.2.3 Resilience-Oriented Service Restoration and Reconfiguration ............................. 11 2.2.4 Nomograms ........................................................................................................... 19 2.3 Consolidated approach ..................................................................................................... 22 3.0 Distribution-level State Estimation and Parameter Estimation ................................................ 25 3.1 Background and Introduction ........................................................................................... 25 3.2 Data and Measurements in Distribution Systems ............................................................. 28 3.3 Distribution System State Estimation............................................................................... 29 3.3.1 Weighted Least Squares Formulation for Distribution System State Estimation . 29 3.3.2 Simulation Studies ................................................................................................. 34 3.3.3 Case 6: IEEE 8500 Node Test System .................................................................. 38 3.4 Distribution System Parameter Estimation ...................................................................... 39 3.4.1 Methods ................................................................................................................. 40 3.4.2 Simulation Studies ................................................................................................. 42 3.5 Conclusions and Future Work .......................................................................................... 47 4.0 Conclusions .............................................................................................................................. 48

A comparative study of distribution system parameter estimation methods

In this paper, we compare two parameter estimation methods for distribution systems: 1) residual sensitivity analysis and 2) state-vector augmentation with a Kalman filter. These two methods were originally proposed for transmission systems, and are still the most commonly used methods for parameter estimation. Distribution systems have much lower measurement redundancy than transmission systems; therefore, estimating parameters is significantly more difficult. To increase the robustness of parameter estimation, the two methods are applied with combined measurement snapshots (measurement sets taken at different points in time) so that the redundancy for computing the parameter values is increased. The advantages and disadvantages of both methods are discussed. The results of this paper show that state-vector augmentation with a Kalman filter is a better approach for parameter estimation in distribution systems. Simulation studies are done on a modified version of IEEE 13-Node Test Feeder with varying levels of measurement noise and non-zero error in the other system model parameters.