An Ultra-short-term Wind Speed Forecasting Model Based On Fluctuation Process Pattern Recognition

With the vigorous development of wind power generation,the global installed capacity of wind power has increased significantly,and its penetration rate in the power grid has gradually increased.However,the volatility,intermittence and randomness of wind power are increasingly detrimental to the stability of the power system.Wind speed is the main factor causing wind power fluctuations,and the accuracy of wind speed forecasting is related to the accuracy of wind power prediction.Therefore,improving the accuracy of wind speed forecasting is also one of the most effective means to promote the consumption of wind power.Against this background,this paper has conducted in-depth research on the ultra-short-term forecasting of wind speed.Most existing researches all regard historical wind speed as a continuous and uninterrupted sequence,and ignore it contains different fluctuation processes.Moreover,the selected training samples have low compatibility with the test samples,which severely limit the forecasting accuracy.To improve the accuracy of wind speed and wind power forecasting,an ultra-short-term wind speed forecasting model based on time-scale information classification and dynamic adaptive modeling is proposed in this paper.The model consists of four steps.Firstly,since there may be noise in wind speed data,this paper uses wavelet analysis for noise reduction to obtain a relatively stable signal.Secondly,a series of wind processes are extracted from the historical wind speed sequence,and each wind process describes a variation of wind speed from rising to falling.Thirdly,as the fluctuation law of wind speed is closely related to the duration of the wind process,the wind processes are classified into two time scales by analyzing the distribution of wind process duration,one is the short time scale wind process,the other is the long time scale wind process.The former’s wind speed is likely to show a downward trend within the forecasted time and the latter’s wind speed is likely to show a continuous fluctuation trend within the forecasted time.And the wind process time scale classification model is established to preliminarily judge the future fluctuation law of wind speed.Finally,based on the recognition result of the current wind process time scale and the high compatible training samples screened out for current approach input by applying the complex network,the corresponding classification forecasting model is established,which can effectively avoid the large error caused by the single mapping relationship established by the unified model,so as to further improving the forecasting accuracy.In order to verify the validity of the model,the wind speed and meteorological data from the Desert Rock,Nevada,USA,are selected for example verification,and it is found that the proposed model has higher forecasting accuracy than the benchmark model using the same forecasting algorithms but without considering the time scale and data screening based on complex network.