Abnormal Data Discrimination And Forecasting For Wind Power Time Series

The forecasting of wind power is an important component to achieve large-scale integration,the accurate forecasting of wind power has play a vital role to reduce the impact of wind power on the stable operation of power system,and reasonably distribute the electric energy.Owing to the interminably and uncertainty of wind speed,the wind power output performance with the strong volatility and randomness and wind power prediction is not satisfied.Chaos is a phenomenon of inherent randomness in deterministic nonlinear systems,with the development of the nonlinear dynamical system,the dynamical information of chaotic time series is gradually reflected abundantly,its provides a scientific method for the chaotic characteristics analysis and prediction of wind power.Based on Voronoi graph theory and chaos theory,the wind power abnormal data were discriminated and the characteristic of wind power were analyzed and its forecasting were discussed.The main contents are as follows:1.Based on Voronoi diagram,the anomaly data detection method based on the characteristic extraction of piecewise linear representation is established,and the abnormal data of wind power is detected with high detection efficiency.Using the piecewise linear representation of time series to extract features,a new method for outlier detection is established.2.The performance of different methods for determining delay time and embedding dimension are analyzed qualitatively in the phase space reconstruction theory,and an optimal phase space reconstruction method is determined.3.The chaotic characteristic and forecasting of wind power time series are analyzed,the correlation dimension and the largest Lyapunov exponent are calculated quantitative,those index is used to identity the chaotic characteristic of wind power sequence,and the adding-weighted one-rank Local-region multi-steps forecasting model and wavelet neural network are adopted for short-term wind power sequence prediction.Based on the predictable size of chaotic time series,the wavelet neural network prediction method with chaos algorithm is established,and the method has used to wind power sequence,the average absolute error is reduced to 4.7795%.