Search

With rapid growth of wind power integration into the modern power system, Wind Power Prediction (WPP) plays an increasingly important role in the planning and operation of electric power system. However, the wind power time series always exhibits nonlinear and non-stationary characteristics, which is still with great challenge to be predicted accurately. To overcome the challenge, a Stacked Denoising Auto Encoders (SDAE) based deep learning approach for WPP is proposed in this paper. Firstly, SDAE with three hidden layers is designed to capture the nonlinear and complex characteristics from the reference data sequence, and the optimal initial connection weights of the deep neural network is obtained by the layer-wise pre-training process. Secondly, the back propagation algorithm is applied to fine-tune the weights of the whole network. To achieve the optimal network architecture, the Bat Algorithm (BA) is adopted to identify the number of neurons of the hidden layers for each denoising auto encoder (DAE). Finally, the proposed method is evaluated by the data from a real wind farm and compared with the Back-propagation Neural Network (BPNN) and the Support Vector Machine (SVM). The results show that SDAE is with the ability to learn the nonlinear and non-stationary characteristics of wind power data and the WPP Root Mean Square Error (RMSE) are reduced by 3.49% and 1.59% in comparison with BPNN and SVM, respectively, which is applicable for practical applications in electric power system.

Outlier detection is an important task in data mining and has been well studied on precise data. However, outlier detection on uncertain objects is particularly challenging. In this paper, firstly, the conceptions about density-based top-k uncertain outlier detection are defined. Secondly, an algorithm of density-based Top-k outlier detection on uncertain objects is proposed, the time complexity of which is polynomial. Finally, the experiment illustrates the effectiveness and efficiency of the algorithm.

This paper presents an application method of Bayesian networks (BN) to the reliability assessment of power systems. Bayesian networks provide a flexible framework to represent probabilistic information and to make inference on it. Uncertainty and dependency of the components' information in a system are easily incorporated in the analysis. The flexibility of the probabilistic inference algorithms in Bayesian networks permit to compute both the system's reliability indices and the mutual affection on reliability indices of all components. However, a BN cannot be constructed easily based on the topology of the relating power system. The paper gives a new method to construct a Bayesian network based on the assessed system's fault tree or its minimal path set. The method is efficient and can compute components failure probabilities on the condition of the system failure. Its advantages are demonstrated through two examples.