BC Hydro's on-line transient stability assessment (TSA) model development, analysis and post-processing

Convolutional neural network-based power system transient stability assessment and instability mode prediction

Online transient stability assessment and preventive control are important measures to prevent power system blackouts. Based on the deep cross network (DCN), an online assessment and preventive control method for power system transient stability is proposed. The deep cross network is developed to estimate the critical clearing time (CCT) of the credible fault contingencies. Based on the deep cross network model, the sensitivity of the active power of the generator set to the fault critical clearing time is calculated and then the DCN model-embedded transient stability constrained optimal power flow (TSCOPF) model is proposed to compute the preventive generation re-dispatch strategy. This TSCOPF model is computed by successive linear programming assisted by the DCN. The simulation results of the IEEE 39-bus test system show that this method can give early warning to the operation mode with instability risk and generate preventive control strategies, which can effectively improve the safety level of power grid operation.

Online transient stability assessment (TSA) has always been a tough problem for power systems. One of the promising solutions is to extract hidden stability rules from historical data by machine learning algorithms. These algorithms have not been fully accommodated to TSA, since power system has its special characteristics. To ensure conservativeness of TSA, this paper proposes a synthetic stability classifier based on reformed support vector machines. It separates samples into stable, unstable and grey area. The stable and unstable classes are expected to be exactly correct. Moreover, an SVM solver for large scale problem is designed based on sequential minimal optimization (SMO). It decomposes large scale training into parallel small scale training so as to speed up computation. Case studies on IEEE 39-bus system show no false dismissals and demonstrate the advantage of proposed classifier and SVM solver.

An on-line transient stability emergency control approach is proposed, which couples an open-loop and a closed-loop emergency control technique. The open-loop technique uses on-line transient stability assessment in order to adapt the settings of automatic system protection schemes to the current operating conditions. On the other hand, the closed-loop technique uses measurements in order to design and trigger countermeasures, after the contingency has actually happened, then to continue monitoring in a closed-loop fashion. The approach aims at combining advantages of event-based and measurement-based system protection schemes, namely, speed of action and robustness with respect to uncertainties in system modeling. It can also comply with economic criteria.

This paper presents a new candidate method for on-line transient security assessment and enhancement based on Pattern Recognition techniques. Transient state variables are used as primary features and the classifier is obtained in two phases, through the combination of an optimum discriminant transformation together with a probabilistic procedure. The proposed security functions can be further refined by means of a second level classifier based on an optimal discriminant plane and using a weighted K nearest neighbour rule. Finally it is shown that this method provides also an efficient and simple tool for on-line preventive control. Extensive results obtained for several contingencies in the CIGRE test system are presented and discussed.

Deep learning-based transient stability assessment framework for large-scale modern power system

Areas discussed are power system modeling, state estimation, structure decomposition, state forecasting, clustering and security measure development. A detailed dynamic model of a multi-machine power system has been developed. A process state estimator was developed to estimate the long-term dynamic behavior of the power system. The algorithm is identical to the extended Kalman filter but has a modified process noise driving term. A two-stage structure estimation technique was proposed for identifying the power system network configuration. Two approaches to structure decomposition were investigated. A time-scale decomposition of the system equations, based on a singular perturbation approach, was evaluated using a detailed model of a generating system. Spatial decomposition was examined by applying an optimal network decomposition technique to a 39-bus test system. Stochastic approximation based approaches to estimator simplification were examined. Explicit expressions were obtained for the evolution of the first and second moments of the system state. Research into security measures proceeded in three directions. The first area involves viewing the security assessment problem as a hyperplane crossing problem for a stochastic process. The second approach examined the stability of an unforced linear system where the system coefficients are subject to future jumps. The third area of research has led to the formulation of a security measure suitable for on-line assessment of transient stability.

The method of generalized equal-area criterion (GEAC) for transient stability analysis of power systems and its recent generalization to cover a wider range of faults are explained. The method is based on an exact mathematical transformation of the time-invariant dynamical equations of an n-machine system into an equivalent single-machine-infinite-bus (SMIB) system equation with time-varying parameters. Further approximations are made to assign appropriate constant values to the SMIB parameters, thus enabling the use of a modified version of the well-known equal-area criterion. An approximate stability index is then defined to yield the degree of stability/instability. GEAC and time simulation results showed excellent agreement in the majority of cases; in the worst case the difference did not exceed 7%. Some tests results are presented. Potential applications of the GEAC method for online transient stability assessment are discussed. >

Transient stability greatly challenges the safe and stable operation of the power system. Therefore, it is extremely necessary to assessment transient stability quickly and accurately. This paper proposes an intelligent model for transient stability assessment (TSA) based on the separation feature and deep belief networks (DBN) for large-scale AC/DC hybrid power grid. With this model, a training method which is suitable for large-scale AC/DC hybrid power grids was proposed. According to the correlation degrees between different electrical characteristics and power system transient stability, the input features are separated into operation mode and disturbance mode, and then, using DBN to extract the features of the two types of input data respectively. In order to adapt the large-scale AC/DC hybrid power grid, different pre-training sample sets are applied for different DBN feature extraction networks to improve the generalization ability and reduce the complexity of model training. Test results with the proposed model on the 265-bus AC/DC hybrid power grid designed by the China Electric Power Research Institute indicated that the new model is easy to train, and it can effectively reduce the influence of imbalanced samples, with good generalization ability.

Transient stability assessment (TSA) based on machine learning techniques is a classical problem of classification. In this paper, a new transient stability assessment model using the weighted random forest (WRF) is proposed. Random forest, which runs fast and can scale to very high dimensional datasets, is one of the most popular methods applied to TSA after faults occurred. However, in the actual operation of modern power system, unstable samples constitute only a very small minority of the cases. Besides, the traditional random forest algorithm does not work very well on imbalanced data. To tackle the problem, weighted random forest is adopted to lean the interest towards to the correct classification of rare class, and improve the prediction performance. Case study on the New England 39-bus test power system exhibits that WRF has better performance than the original random forest.

This article presents the Lyapunov functions based on the artificial intelligence (AI) method for transient stability assessment and the determination of stability region (SR). First, Lyapunov stability theory and the definition of SR are introduced. Then, the characteristics of neural networks as a general function approximator are employed as the Lyapunov function learner, and the Lyapunov function is constructed combined with stochastic gradient descent (SGD). Then, the falsifier's task is to find the state vectors that violate Lyapunov stability conditions, and the counterexamples would be added to the training set for the function learner to accelerate convergence. After obtaining the Lyapunov function of power system, the estimation of SR boundary can be represented by the maximum level set of Lyapunov function. Finally, the IEEE 9-bus 3-machine system is used as test system to demonstrate the validity and effectiveness of the proposed construction method of Lyapunov functions for power system transient stability analysis.

On-line Transient Stability Assessment (TSA) program can fast identify a list of critical contingencies. A real-time critical contingency (RTCC) detection system is to detect, on the basis of the real-time measurements provided by PMUs and the list of critical contingencies, the occurrence of a critical contingency is presented. The RTCC detection system is composed of five detectors. Each of them is designed and placed in a sequential order so as to achieve absolute capture of the occurrence of critical contingencies. The system has been evaluated in 2 test systems and one large-scale power systems with very promising results. The effectiveness of the developed critical contingency detection system is evaluated on 96 contingencies of a 10-machine, 39-bus power system, on 108 contingencies of a 50-generator, 145-bus power system and on 4588 contingencies of a 313-machine, 1648-bus power system.

As part of a joint effort between B.C. Hydro and ALSTOM ESCA, an online transient stability analysis (TSA) capability was developed for addition to a new energy management system being installed at the B.C. Hydro control center. Following the original development, a number of research projects enhanced the original development making it more efficient and generic for applications to other utility systems. The requirement to post available transfer capability for a up to one year ahead, is shifting the paradigm for online DSA. Sections 2 and 3 of this article briefly describe the design and the architecture of an online DSA. Section 4 describes the requirement for ATC calculation and how it will affect the future of online DSA. Finally Section 5 describes the areas of development and improvements for the existing online DSA to meet the challenges of the future.

Maintaining accurate and fast transient stability is essential for safe operation of the power system. With the development of wide-area measurement system, machine learning–based transient stability assessment has become the trend. However, in realistic application of the power system, the impacts on evaluation rules between critical samples and noncritical samples are different. Thus, an improved cost-sensitive coefficient assignment method based on fault severity is proposed. First, the fault severity of each unstable sample is calculated. Then, the correction coefficient of the loss function of the unstable sample is linearized according to different fault severities. The closer the sample is to the critical case, the higher the cost coefficient is. Finally, the improved cost-sensitive method is combined with the deep learning model and tested in the IEEE-39 bus system. As shown in the results, the improved cost-sensitive method, which gives different correction coefficients to samples according to different fault severities, has better performance.

Traditional machine learning models are subject to changes in the data distribution. Once the topology of power grid changes, the distribution of power grid data will change, resulting in the consequence that the former transient stability assessment model based on machine learning is no longer applicable and needs to be retrained. However, the power network after the topology changes cannot provide enough data for the training of the new model. To solve the problem above, this paper proposes a method of power system transient stability assessment based on transfer learning, which fully taps the potential of past data and assists a small amount of newly acquired data in training the model. The simulation results on the CEPRI36 system show that the proposed method can effectively improve the accuracy of original classification models, especially for weak classifiers with insufficient generalization ability. The disadvantage of transfer learning lies in the long training time, which still needs to be improved.