Tingyan Guo

Probabilistic Framework for Assessing the Accuracy of Data Mining Tool for Online Prediction of Transient Stability

The paper presents a generic probabilistic framework for assessing the accuracy of online prediction of power system transient stability based on phasor measurement unit (PMU) measurements and data mining techniques. It allows fair comparison of different data mining models in terms of the accuracy of the prediction. To illustrate the concept, a decision tree (DT) method is used as an example of a data mining technique. It is implemented in a 16-machine, 68-bus test power system. Generator rotor angles and speeds provided by PMUs during post-fault condition are chosen as predictors. The performance of the DT based prediction method is tested using a wide variety of disturbances with probabilistically modeled locations, durations, types of fault and the system loading levels. The accuracy of prediction is approximately 98.5% immediately following the fault clearance and can increase to almost 100% if the prediction is made 2.5 s after the fault clearance.

Online Identification of Power System Dynamic Signature Using PMU Measurements and Data Mining

This paper proposes a two-stage methodology for online identification of power system dynamic signature using phasor measurement unit (PMU) measurements and data mining. Only transient stability status is usually predicted in the literature to assist with corrective control, without the dynamic behavior of generators in the event of instability. This paper uses traditional binary classification to identify transient stability in the first stage, and then develops a novel methodology to predict the nature of unstable dynamic behavior in the second stage. The method firstly applies hierarchical clustering to define patterns of unstable dynamic behavior of generators, and then applies different multiclass classification techniques, including decision tree, ensemble decision tree and multiclass support vector machine to identify characterized unstable responses. The proposed methodology is demonstrated on a multi-area transmission test system. High prediction accuracy at both stages of identification is demonstrated.

The effect of quality and availability of measurement signals on accuracy of on-line prediction of transient stability using decision tree method

This paper investigates the practical issues associated with on-line prediction of transient stability using decision tree (DT) method. The issues of quality and availability of the measurement signals provided by wide area measurement system (WAMS) are discussed and their effects on the accuracy of performance of the DT are evaluated. The surrogate split method included in the classification and regression tree (CART) algorithm is used to handle the unavailability of measurement signals, and noise present in the on-line data is modeled as white Gaussian noise (WGN) with various signal-to-noise ratio (SNR). Case studies are carried out in the 16 machine, 68 bus power system. The results quantify the reduction of the performance of the DT method.

Evaluation of classification methods for on-line identification of power system dynamic signature

The paper investigates the use of Decision Tree (DT), Ensemble DT and multiclass Support Vector Machine (SVM) for on-line prediction of post-fault system dynamic signature based on Phasor Measurement Unit (PMU) measurements. The performance of these multiclass classification techniques is compared in terms of i) how fast the prediction about generator grouping can be made after the clearance of transient disturbance and ii) the accuracy of prediction. The application of these methods is illustrated on a 16-machine, 68-bus test system. Results indicate that the Ensemble DT method performs the best by achieving accuracy of close to 90% using 10 cycles data of post-disturbance generator rotor angles as predictors and over 90% using 30 cycles data of rotor angles as predictors.

Application of sequential testing problem to online detection of transient stability status for power systems

We address the problem of predicting the transient stability status of a power system as quickly as possible in real time subject to probabilistic risk constraints. The goal is to minimise the average time taken after a fault to make the prediction, and the method is based on ideas from statistical sequential analysis. The proposed approach combines probabilistic neural networks with dynamic programming. Simulation results show an approximately three-fold increase in prediction speed when compared to the use of pre-committed (fixed) prediction times.

Comparison of ensemble decision tree methods for on-line identification of power system dynamic signature considering availability of PMU measurements

This paper compares the most commonly used ensemble decision tree methods for on-line identification of power system dynamic signature considering the availability of Phasor Measurement Units (PMU) measurements. Since previous work has shown that the surrogate split method included in classification and regression tree is not good enough to handle the unavailability of measurement signals, more effective methods are needed to be explored. Bagging, boosting and random forest methods are investigated and compared in this work. When evaluating their performance, all possible scenarios of missing PMU measurements are tested for the test network. For each ensemble decision tree model, the result is presented as a probabilistic classification error depending on the availability of PMU signals. The test network used is the 16-machine, 68-bus reduced order equivalent model of the New England Test System and the New York Power System.

Global assessment of power quality performance of networks using the analytic hierarchy process model

Power Quality is one of the critical issues when operating contemporary distribution networks. The increased interest in PQ is due to the increased employment of sensitive equipment loads and renewable generation technologies. Currently, the evaluation of PQ network performance is based on evaluating different phenomena separately, and a standardised way to evaluate the PQ as a whole for a bus or a network is yet to be applied or widely accepted. The paper presents a methodology for combining the performances of different PQ phenomena and expressing the PQ performance of a bus using a single index. The proposed methodology adopts an Analytic Hierarchy Process (AHP) model to combine the harmonics, unbalance and voltage sag performances in one proposed index; i.e. Compound Bus PQ Index (CBPQI). The separate and cumulative PQ performances of a 295-bus generic distribution network were evaluated, compared and demonstrated using heat maps.

Probabilistic Framework for Online Identification of Dynamic Behavior of Power Systems With Renewable Generation

The paper introduces a probabilistic framework for online identification of post fault dynamic behavior of power systems with renewable generation. The framework is based on decision trees and hierarchical clustering and incorporates uncertainties associated with network operating conditions, topology changes, faults, and renewable generation. In addition to identifying unstable generator groups, the developed clustering methodology also facilitates identification of the sequence in which the groups lose synchronism. The framework is illustrated on a modified version of the IEEE 68 bus test network incorporating significant portion of renewable generation.

Impact of measurement signals on the accuracy of online identification of power system dynamic signature

In order to identify the dynamic signature of power systems, it is important to monitor the generator rotor angles. The rotor angles can be obtained either by direct measurements of the rotor position or by indirect calculation of rotor angle from voltage and current measurements. Considering limited availability of direct measurements of rotor angles the indirect method is used in this paper taking into account the errors in calculated rotor angles. Decision tree based algorithm is used afterwards for online identification of power system dynamic signatures. Finally, the impact of the accuracy of measurement signals on the accuracy of the assessment of system dynamic signature is discussed.

Probabilistic assessment of voltage control zones and visualization using choropleth map

Voltage control zones are commonly used within power systems, for applications like Secondary Voltage Control and localized reactive power markets. The idea is to partition a power system into zones that are weakly coupled in terms of voltage control. Conventionally, these control zones are defined with a deterministic approach and are fixed. However, recent research has demonstrated that they may be altered due to variation of the network structure and therefore suggested on-line reconfiguration of the zones. This paper goes a further step to probabilistically asses how the zoning could be affected by the uncertainties of operating condition (i.e., how the zones vary within a certain amount of time). Forced network outage, as the uncertain factor that affects the electrical distance between buses, is modeled in a realistic manner. Choropleth map is used to visualize the results. The 39-bus New England Test System (NETS) is used for demonstration.