Research Reports filed with TEPCO

Dynamic contingency screening is a fundamental function of an on-line TSA (transient stability analysis) system. Its main objective is to identify contingencies that are definitely stable and thereby avoid further stability analysis for these contingencies. The overall computational speed of an on-line TSA system depends greatly on the effectiveness of the dynamic contingency screening. The strategy of using an effective scheme to screen out a large number of stable contingencies and to apply detailed simulation programs only to potentially unstable contingencies is well recognized and has been successfully implemented in on-line static security assessment.

Several research developments of on-line dynamic contingency screening have been reported in the literature. The methods behind these developments can be categorized into two classes: the artificial intelligence (AI) based approach and the direct method based approach. The AI based approach all first perform extensive off-line numerical simulations in order to capture the essential stability features of the system dynamic behavior; then they construct a classifier attempting to correctly classify new, unseen on -line contingencies. As such, the AI approach is likely to become ineffective for on-line application to current or near-future power systems because little correlation may exist between on-line operational data and presumed off-line analysis data.

The following features are essential for any candidate (classifier) intended for on-line dynamic contingency screening of current or near future power systems:

- (reliability measure) absolute capture of unstable contingencies; i.e. no unstable (single-swing or multi-swing) contingencies are missed. In other words, the ratio of the captured unstable contingencies to the actual critical contingencies is 1.

- (efficiency measure) high yield of screening out stable contingencies, i.e. the ratio of the stable contingencies detected to the actual stable contingency cases is as close to 1 as possible.

- (on-line computation) little need of off-line computations and/or adjustments in order to meet with the constantly changing and uncertain operating conditions.

- (speed measure) high speed, i.e. fast classification for each contingency case.

- (performance measure) robust performance with respect to changes in power system operating conditions.

My research team and TEPCO (Tokyo Electric Power Company, Tokyo, Japan) have developed a sequence of improved BCU classifiers for on-line dynamic contingency screening to meet the above five requirement. The sequence of improved BCU classifiers, which recently awarded a U.S. patent, will perform on-line transient stability classification in a particular sequential order such that each classifier screens out not only stable and unstable contingencies but also those contingencies which may cause degraded performance for the classifiers that follow sequentially. The classification process of the BCU classifiers is an algorithmic procedure. As such, the scenario of (N-2) dynamic contingency screening can be effectively handled by the proposed BCU classifiers while they are not easily handled by other approaches This points out the vital role an algorithmic procedure can play in developing an on-line dynamic contingency screening scheme for large-scale power systems.

Two-cycle on-line contingency screening

Due to heavy computational efforts required for look-ahead dynamic contingency screening/ranking, we proposed a 2-by-2 strategy which examines dynamic security constraints of a list of credible contingencies in two different time-scale cycles in which there are two stages of tasks in each cycle. There two cycles are 15-minute cycle and one-hour cycle. The overall purpose of the hour-ahead cycle is to assure a comprehensive dynamic security assessment of a credible contingency list within the look-ahead horizon of the next 60 minutes. It includes the identification of the critical system operating limits and of critical contingencies to be carefully monitored and analyzed in the 15-min cycles. In this cycle a relatively large amount of CPU time is available for computational tasks. This allows for processing contingencies in a considerably large geographical portion of power systems over an extended time period. Outputs of this cycle include a short list of unstable and critical contingencies for further detailed analysis in the 15-min cycle. Hence, it is necessary to design effective architectures of two time-scale cycles for performing look-ahead dynamic contingency screening and ranking.

The synchronization of the two cycles will be as follows. The one-hour cycle starts one hour early and load with the loading condition in the next hour. The one-hour cycle will conduction a comprehensive screening against supplied 3000 contingencies. At the end of the hour, a short list of 200 insecure, critical contingencies and other information regarding the rest 2500 contingencies will be outputted to the 15-minutes cycle. The 15-minute cycle update the short list base the information from the one-hour cycle and the loading condition in the next 15 minutes. The final output of the contingency screening subsystem will be a list of 500 contingencies, which include the unstable contingency, critical contingencies, and the so-called undecided contingencies for which TEPCO-BCU cannot offer an accurate assessment because certain mathematical conditions are not satisfied. The hour-cycle will keep running hour after hour. The 15-minute cycle will keep running every 15 minutes except the very first hour. While the hour cycle will need one-hour to complete, the 15-minute cycle only needs 5 minutes to update the short list, leaving 10 minutes for the detailed analysis.

Modeling and contingency support

- Generator: classical to two-axis 6^th order modals

- Excitation system: all IEEE standard exciter/AVR and PSS models and common extended models

- Load: ZIP model, voltage dependent model, discharge lighting model

- SVC

- User-defined modeling (transfer function)

- Speed governing system

- Load: induction motor

- Other FACTS models and HVDC model.

- Under-voltage/frequency load shedding, switchable shunts, distance relay

Research Reports:

1998

- Enhancements of BCU Program with TEPCO Transient Stability Models

- Research into BCU Method for Practical Application to Comprehensive Stability Model

- Extensions of BCU Method to Transient Stability Models with Non-smooth Load Models

1999

- Development of a Group-based BCU Method ¡V Part I: Research

- Development of Improved BCU Classifier for TEPCO Incorporated Analytical System

- Study of the Applicability of Improved BCU Classifiers for Multi-swing Stability Analysis

- Continual Development of BCU Classifiers (Version 2)

2000

- Feasibility Study of Developing Screening Methods to Decide Network Reconfiguration and Network Reloading for Maintaining/Improving Transient Stability

- Feasibility Studies of Developing Time-Domain Energy Indices for Dynamic Security Assessment

- Development of a group-based BCU Classifier -- Part II: Development

2001

- Study of the precision improvement for the Group-based BCU Method

- Feasibility Study of Developing New Time-Domain Energy Indices for TEPCO Power System

2002

- Study of Detailed Excitation Models in BCU Program for TEPCO Power System

2003

- Development and Implementation of Models of Generator Controllers and Phase-shifters for BCU and GBCU Programs

- Improvement in the Performance of Group-based BCU Programs

2004

- Development and Implementation of Group-based BCU Program

- Study on Computing Method of Energy Margin Index for BCU and Group-based BCU Methods

2005

- Group-based BCU Methodology for On-Line Dynamical Security Assessments and Energy Margin Calculations of Practical Power Systems

- Implementation of Additional Contingency Options for DSA System by BCU Method

- Research on Prototype Algorithm for Transient Stability Control System

2006

- Excitation Systems Equivalencing and Generator Aggregation for Transient Stability Analysis and Control

- Improved TEPCO-BCU and BCU Screening Classifiers for On-line Dynamic Security Assessment of Large-Scale Power Systems

- Development of Parallel TEPCO-BCU for On-line Dynamic Security Assessment III

- Application of Parallel TEPCO-BCU for On-line Dynamic Security Assessment of PJM Online Test System