Study Of Short-Term Power Forecasting In Coastal Wind Farm

Under the new situations and circumstances of a worldwide avocation of energy conservation and reduction of pollutant emission, the wind energy, as the representative of the renewable energy sources has caused extensive concern. China has pretty rich wind resources and wind power industry has developed rapidly in recent years. As one research field of wind-electric technology, the short-term forecast of wind power field, which plays an important role in the stable operation and the proper modulation of wind power network, gets more compelling due to the special geographical conditions and the wind speed characteristics of wind power field. Although the traditional methods of forecast have their own merits in the short-term forecast of wind power field, they couldn't be a good application in the coastal wind farms. The content of the traditional research covers a limited area of the uncertainty analysis on the prediction results of power. In addition, prediction results not well reflect the value of their scientific guidance. Therefore, the main contents of this paper are to build new power prediction model and make uncertainty analysis of power prediction results.For the difficult forecasting problems of non-linear, non-stationary power and wind speed series, this paper proposes a short-term power prediction model based on improved Empirical Mode Decomposition (EMD) and Artificial Neural Network (ANN). Power series can be decomposed into different series using the improved fitting algorithm in the process of fitting the envelope, in order to reduce the sequence of non-stationary, and then ANN is used to forecast power by using each component. The simulation example shows that the model has higher prediction accuracy.For the difficult problems of measuring values interfered by a number of external factors and the nature of the interaction between their, this paper proposes a short-term power prediction model based on Independent Component Analysis (ICA) and Least Squares Support Vector Machine (LS-SVM). ICA is applied to power forecasting to ensure that the extracted components are independent of each, and then each component is predicted using LS-SVM, the final results obtained by modifying the preliminary predicting power according to existing information. The simulation example shows that this model has higher accuracy, rationality and effectiveness, compared with other methods.This paper makes uncertainty analysis of power prediction results, and proposes a new model to determine the confidence probability based on ICA and conditional probability theory. For the shortcomings of traditional methods that determine the confidence probability, the power independent influence events set can be obtained from ICA, then the problem of determining the confidence probability will transform into the conditional probability and unconditional probability problem whose objects are the power independent influence events. The model is clear and easy, which is fully takes into account occurrence conditions of the target power and the original content. The simulation example shows the confidence probability result has realistic sense, making the study of short-term power forecasting in coastal wind farm more perfect.