Predictability Of Ultra-short Term Wind Power Prediction In Large Scale Wind Farm

As a kind of natural energy, wind energy is inexhaustible in supply and can’t be used out. It is not need for a large number of mining and transportation cost and has the characteristics of zero pollution. It is a kind of typical cheap and clean energy which could sustainable cycle and regeneration. But it is well known that the wind is random and intermittent. The available power supply produced by it is unknown. Severe challenge to the power system will be appeared when a large-scale wind power is allowed to connect to the grid. Wind power prediction plays an important role in the response to these challenges. If the wind power can be accurately predicted, the electric power dispatching department will be able to arrange the scheduling plan reasonably, guarantee the power balance between supply and demand, reduce the influence to the power system effectively when the wind farms connect to the grid, and reduce the spinning reserve capacity and power cost. At present, the study of wind power prediction is mainly focused on the improvement of the forecasting method, but no matter how sophisticated forecasting method is, it can not achieve 100% accurate predictions. Therefore, in view of ultra-short term wind power prediction, in order to do a comprehensive study for wind power prediction, wind power fluctuation under different sampling intervals, multi-step rolling prediction method and the predictability of wind power will be researched in the paper.At first, in the view of different sampling intervals, the fluctuation of wind power sequence is analyzed. In order to evaluate the losing amount of the wind power fluctuations information when the sampling interval changes, fluctuations standard deviation and information entropy are established in the text. Multi-step rolling prediction mode and the relationship between the multi-step rolling prediction error and the information entropy are analyzed, and the strong positive correlation between them is shown in the study case. In the end, the method how to determine a reasonable recommended value of sampling interval is given under the wind power multi-step rolling prediction mode.Secondly, the methods of ultra-short term multi-step wind power prediction are researched. The ANFIS(adaptive neuro-fuzzy inference system) prediction model is established. The subtraction clustering algorithm is used when it form initial fuzzy inference system structure. This algorithm is effective to avoid the combination explosion problem of the artificial setting structure method. The real-time multi-step rolling results of wind power based on the linear regression method, the moving average method and the persistence method are compared with the prediction results based on the proposed ANFIS prediction method. The result of case study shows that the prediction accuracy of the latter method is the highest, and it further illustrates the validity of the ANFIS model.At last, the predictability of wind power sequence is analyzed. Wind power prediction accuracy is not only related to the prediction method adopted, but also associated with degree of predictability of wind power itself. The necessity of the probability in the wind power sequence is explained in this part. To depict the probability of the occurrence of a new volatility mode, based on phase space reconstruction, a recurrence plot and the recurrence rate is proposed to qualitatively and quantitatively depict the probability. The changing rules of the recurrence plot and recurrence rate at different spatial scales were discussed. Based on this, a method to analyze the relationship between the probability of a wind power sequence and prediction error is established. Finally a fair method for evaluating predictions accuracy is provided based on the recurrence rate.