Data mining and graph theory focused solutions to Smart Grid challenges

This paper presents a novel strategy based on pattern discovery algorithm and extended A-Nearest Neighbor (k- NN) classifier to design the power system transient stability assessment (TSA) scheme. A pattern discovery algorithm based on residual analysis and recursive partitioning is employed to identify the latent structure (called pattern) of data samples. An extended A-NN classifier is then designed to label the input sample with stability levels where the identified patterns are applied as initial classified data set. In order to improve the pattern discovery efficiency, a feature selection process is introduced in the first stage to extract kernel-feature subset from the pre-contingency steady state parameters as the input features for the TSA scheme. Applications of the proposed scheme on two IEEE test systems prove its feasibility and show good performance in stability level classifications.

A data mining technique based on pattern discovery and k-nearest neighbor classifier for transient stability assessment

A number of characteristics of power system is hidden in the operating point space, including the stability aspect. This paper presents a data-structure analysis of the operating point space scheme for power system transient stability assessment (TSA) with the adoption of RELIEF, which is an attribute estimation method capable of evaluating the ability of attributes to discriminate in different classes. The stability characteristic of power system can be approximated by those attributes with high RELIEF weights, and the TSA can then be converted into a classification process in the approximate operating point space. By testing on New England 39-bus system, we demonstrate the scheme can effectively approximate the stability characteristic of power system and perform a highly accurate transient stability assessment.

Transient stability assessment based on data-structure analysis of operating point space

A robust operation rule extracting algorithm is proposed based on adaptive time series association analysis here. It is applied to extract association knowledge among the accumulating massive operation data. A data model describing the system state and topological structure before/after fault happening is built up at first. The initial abstract features are independent of the scale of a power system. A recursive principal component analysis method is employed to yield the reduced-size feature space. The extract features are then fed into the associative analysis. For considering temporal constraint, the generated associative rules not only reflect the relationship between power system operating condition and transient stability, but also reveal some valuable information about operating characteristics. Simulation results on on IEEE 39-bus test system demonstrate the feasibility and efficiency of the proposed method.

Operation rule extracting based on time series association analysis in transient stability study

In terms of monitored operational information, a visual transient stability assessment (TSA) approach is presented to power system. In the procedure, the initial high-dimensional operation data points are projected into an intuitively understandable graphical representation with two-dimensional visualization. Moreover, the data with different operation states are divided into two clusters. One is the stable cluster, and the other is the unstable cluster. Consequently, the operation points and the stability boundary can be strongly apparent on the two-dimensional displays after the mapping. In order to improve the performance of visualization, some data compression techniques are applied in the paper. To an unseen operation point, its status could be decided easily and quickly according to the boundary and its position in the projected space. The proposed method is designed to be applicable for any power system. The assessments of the transient stability for the IEEE 10-machine 39-bus system and a real large-scale power system, 4267-bus State Grid of China system, are used to demonstrate the validity of the approach.

Transient Stability Boundary Visualization for Power System

One of the most challenging problems in real-time operation of power system is the assessment of transient stability. Fast and accurate techniques are imperative to achieve on-line transient stability assessment (TSA). Based on the statistical learning theory, a novel learning-based nonlinear classifier, i.e., the support vector machines (SVMs) for TSA was presented here. In the approach, the feature variables, which describe the system state before and after the occurrence of a fault, were selected for TSA. Abundance of initial data was preprocessed by feature extraction to improve the data quality. By using SVM training, models were built and used to predict the operation state whether is stable or not for given operation data. The validity of the approach was verified by the simulation for the 4933-bus state grid of China system

Fast Transient Stability Assessment Based on Data Mining for Large-Scale Power System

While continuously searching for new transient stability assessment (TSA) algorithms based on pattern recognition in recent years, less work was focused on data preprocessing which is one of important TSA components. In general, abundance of original data would increase the complexity of TSA, resulting in a longtime computation. In order to improve the performance of TSA, two novel data preprocessing approaches were proposed in the paper. One is fuzzy ISODATA, and the other is the modified recursive principal component analysis. In these ways, the size of data can be obviously reduced without jeopardizing the results of TSA. Simulation of a 4933-bus state grid of China system indicates higher efficiency in comparison with the traditional TSA

Zhongxi Wu

Papers

Fast Transient Stability Assessment Based on Data Mining for Large-Scale Power System

Novel Approaches to Data Preprocessing For Transient Stability Assessment