The intermittency and volatility of wind power (WP) have restricted the large-scale integration of wind turbines into power systems. Therefore, many researches on the deterministic and probabilistic forecasting of WP are being more important for power system planning and operation. In this work, a deterministic and probabilistic interval prediction for wind farm based on variational mode decomposition (VMD) and weighted least squares support vector machine (LS-SVM) were suggested. VMD is proposed to handle the variability of the novel WP series. In order to overcome the influence on outliers and non-Gaussian error distributions, a weighted LS-SVM is adopted to build deterministic prediction model for WP. Besides, lower upper bound estimation (LUBE) method for probabilistic interval forecasting is presented to quantify the potential risks of the WP series. The LUBE of the optimal prediction intervals is calculated. In the comparative experiments, prediction intervals coverage probability (PICP), prediction intervals normalized average width (PINAW) and normalized average deviation (NAD) are demonstrated to appraise the probabilistic prediction of WP. The simulation results show that the proposed method has much greater performance than other methods.

Wind turbines (WTs) usually work in harsh environments and complex operating conditions, it leads to operating data of WTs with non-stationary, randomness and outlier. These imperfect data will inevitably impact the fault diagnosis of WTs. In this work, a fault diagnosis method for WT generators using normal behavior model (NBM) based on multi-task learning (MTL) is proposed. The imperfect operating data come from WTs are preprocessed and divided into different conditions. Each divided condition is constructed as a task, and a special deep neural network (DNN) is employed to train each task. Different tasks share some parameters through $L_2$ regularization at the bottom layer, and are designed independently at the upper layer. Multiple tasks learn at the same time and promote each other’s performance. We employ actual operational data to test the validity for NBM of WTs. The comparative results indicate that the proposed method overcomes the deficiency of imperfect data and is suitable for fault diagnosis under complex operating conditions Note to Practitioners —Traditional fault diagnosis methods do not divide imperfect data under complex operating conditions, resulting in poor generalization and robustness in practical engineering applications. This work proposes a data-driven fault diagnosis method, which regards the imperfect data under each operating condition as a task and employs MTL to improve the performance of each task. Experiments results show that the proposed method is feasible, but the method has not been tested on other types of wind turbines. Next, we will extend this method to other mechanical components under complex operating conditions. At the same time, we also consider embedding the proposed algorithm into SCADA system as an Active X control object, which can automatically analyze the imperfect data.

The quality index model in slashing process is difficult to build by reason of the outliers and noise data from original data. To the above problem, a fuzzy neural network based on non-Euclidean distance clustering is proposed in which the input space is partitioned into many local regions by the fuzzy clustering based on non-Euclidean distance so that the computation complexity is decreased, and fuzzy rule number is determined by validity function based on both the separation and the compactness among clusterings. Then, the premise parameters and consequent parameters are trained by hybrid learning algorithm. The parameters identification is realized; meanwhile the convergence condition of consequent parameters is obtained by Lyapunov function. Finally, the proposed method is applied to build the quality index model in slashing process in which the experimental data come from the actual slashing process. The experiment results show that the proposed fuzzy neural network for quality index model has lower computation complexity and faster convergence time, comparing with GP-FNN, BPNN, and RBFNN.