Short-term wind speed prediction method based on multi-view wind speed model mining

Abstract

The invention discloses a short-term wind speed prediction method based on multi-view wind speed model mining, which includes: collecting historical wind speed data of a wind farm to form a time series of historical wind speed; determining the delay of the time series by mutual information method; Convert the wind speed time series into matrix data to obtain wind speed samples; extract the characteristic information describing the change law of wind speed from the three perspectives of wind speed statistical information, wind speed change trend and wind speed fluctuation trend according to the wind speed samples; normalize the wind speed samples , to obtain the characteristic information of the normalized wind speed change; based on the characteristic information of the normalized wind speed change, the wind speed samples are clustered through the multi-view clustering algorithm, so as to establish a short-term wind speed prediction model with k clusters Model; calculate the Euclidean distance between the wind speed sample to be predicted and the k clusters determined above, and use the SVR wind speed prediction model corresponding to the cluster with the smallest Euclidean distance to complete the wind speed prediction.

Description

technical field [0001] The invention relates to a short-term wind speed prediction method based on a multi-view wind speed change mode of a wind farm, belonging to the field of electric power information technology Background technique [0002] The energy crisis and environmental crisis caused by the excessive consumption of fossil fuels are becoming more and more serious, and wind power has been rapidly developed as a clean energy source. However, the accurate prediction of wind speed and wind power restricts the stability of large-scale wind power grid-connected power generation. At present, the widely used prediction methods are numerical weather prediction (NWP) and data-driven modeling (data-driven modeling). Numerical forecasting based on meteorological formulas can obtain relatively stable meteorological forecasts, but its calculation costs are high, and the short-term and ultra-short-term wind speed forecasts have large errors. Data-driven modeling is widely used in...

AI Technical Summary

Problems solved by technology

[0004] However, the existing clustering methods mainly use the statistical characteristics of wind speed, such as mean value, variance, etc., or simply concatenate the statistical information and wind speed trend information directly into a single feature information, and it is difficult to combine the wind speed trend information or Statistical information as a feature of an independent whole for pattern clustering analysis

Images