The Correction of Forecast Wind Speed in a Wind Farm Based on Partitioning of the High Correlation of Wind Speed

In order to improve precision and accuracy of wind speed forecast and wind power prediction, and taking unstable factors of observations on the wind turbines into account, a refined and partitioned correction model is established for improving the quality of wind speed forecast on turbine hub height. The wind farm A, which is located in the middle of Inner Mongolia of China, is selected as the target area. Fine analysis on gradient observation, the terrain and the layout of the wind turbines is carried out, and the hourly quality-controlled wind data on the wind tower and the turbine hub height are comprehensive compared, considering the spatial-temporal correlation, deviation of wind tower and turbine wind speed, seasonal influences and wind directions (NW, SW, NE, SE). Wind turbine groups are partitioned by wind speed subsequently in wind farm A. Characteristics of wind speed of wind tower and groups that are affected by turbulence, wake flow and wake superposition are studied. Using the method of Kalman filter, through direct (from numerical model to turbine groups directly) and indirect (from numerical model to the wind tower firstly, and then to the turbine groups) correction schemes, wind speed correction is done, respectively.Results show that the division of wind turbine groups in every wind speed is high correlative. The indirect correction of turbine groups using the gradient observation of the wind tower can correct wind speed forecast more effectively than model forecast and direct correction. Correlation (R) and the absolute value of error (E) between corrected wind speed of indirect correction and observed wind speed are improved by different degrees in each wind turbine group. R increases from 0.18-0.72 to 0.67-0.91, and E is reduced from 1.6-2.9 m·s^-1 to 1.0-1.5 m·s^-1 after indirect correction, compared to R increasing to 0.39-0.87 and E decreasing to 1.2-2.1 m·s^-1 after direct correction. In addition, the indirect correction scheme plays a very good role in controlling the E of the low velocity section ( < 3 m·s^-1) and the high speed section (>9 m·s^-1). The indirect correction scheme performs better when the difference between forecast wind speed and observed wind speed are relatively larger. Therefore, this method has is a good reference for improving prediction precision and accuracy, and it should be tested and spread to more wind farms in the service.