Short-term probabilistic wind power forecasting can provide critical quantified uncertainty information of wind generation for power system operation and control. It would be difficult to develop a universal forecasting model dominating over other alternative models because of the inherent stochastic nature of wind power. Therefore, a novel multi-model combination (MMC) approach for short-term probabilistic wind generation forecasting is proposed in this paper to exploit the advantages of different forecasting models. The proposed approach can combine different forecasting models those provide different kinds of probability density functions to improve the probabilistic forecast accuracy. Three probabilistic forecasting models based on the sparse Bayesian learning, kernel density estimation and beta distribution fitting are used to form the combined model. The parameters of the MMC model are solved by two-step optimization. Numerical tests illustrate the effectiveness of the proposed MMC approach.

The active distribution network has witnessed an increasing penetration of distributed generation (DG) while the stochasticity and variability arising from DGs also impose significant challenges on system operation. To mightily accommodate the uncertainty of DG, we introduce a distributionally robust chance-constrained dynamic reconfiguration approach for a three-phase unbalanced distribution network. The proposed framework optimizes the switching cost and the expected power supply cost from upstream grid, and stipulates that the chance constraints hold under the worst-case distribution within a novel ambiguity set, which incorporates the Wasserstein distance and the first-order moment. Then we develop tractable and scalable solution methods to tackle the expected objective function and chance constraints. As a result, the proposed model is reduced to a mixed-integer linear programming problem that can readily be implemented. Numerical experiments are carried out on the IEEE 34-bus and 123-bus test systems to demonstrate the effectiveness and efficiency of the suggested approach.

This paper proposes a non-parametric probabilistic wind turbine power forecast (WTPF) approach based on the empirical dynamic modeling. According to Takens' theorem, the approach applies historical wind turbine power (WTP) time series to form the reconstructed state space. The moving trajectories of state vectors in the space reflect the evolution laws of WTP, and the divergence of adjacent trajectories indicates the uncertainty existing in the forecast. An enhanced simplex projection algorithm is developed to forecast the probability distribution as well as the prediction interval of WTP with respect to the trajectories. Without any assumptions on the distribution types or physical equations, the proposed approach can overcome the drawbacks of distribution type misspecification or physical model incorrectness in the conventional forecast approaches. The real-time updated state vector set in the reconstructed space further merits the reliability and flexibility of the proposed approach. Case studies employ the proposed approach as well as several benchmarks to achieve 5-min, 15-min and 30-min ahead probabilistic WTPFs of Penglai wind farm located in Eastern China, and the superior performance demonstrates the effectiveness of the proposed approach.

A probabilistic deep autoencoder is proposed to reconstruct power system measurements in this paper, which can be utilized in outlier detection and reconstruction. A nonparametric distribution estimation method is employed to capture the uncertainty information of the measured data. The estimated confidence intervals of the measured data are extracted from the estimated distribution and used as input to the first layer of neural networks. Through multilayer encoding and decoding processes, the intervals of measurements are reconstructed, which are further applied to detect and replace outliers. Simulation results verify the effectiveness of the proposed method.

The strong volatility of load and intermittent generations may lead to the tension of operating reserve in the modern power system. However, the conventional operating reserve providers, such as coal-fired generating units, are costly and not environment friendly. To address the growing needs for operating reserve, the day-ahead market for operating reserve have been launched in practice, such as Midwest ISO of the US. The high penetration of renewables has contributed to the increasing demand for the flexible dispatch in the generation side. Renewable generation, such as wind power, is being allowed to participate in the markets by providing auxiliary service. Therefore, wind power can be utilized in a positive way by providing operating reserve. A temporal and spatial multi-state model is developed to exploit the potentiality of wind power for providing operating reserve in this paper. First, the multi-state model for different wind farms is formulated based on the predictive distribution of wind power and correlations among different time periods and different wind farms. Second, the upward and downward reserve capacity of wind power is analyzed quantitatively based on the proposed multi-state model. Then, the adjustable reserve capacity of wind power is utilized in a stochastic day-ahead unit commitment (DAUC) model to determine the optimal operating reserve schedule by balancing the generation economy and wind power dispatch margin. The forecast day-ahead expectations and the possible real-time variations of wind power are considered in the proposed DAUC model. Illustrative results demonstrate the effectiveness of the proposed model.