Analysis Of The Wind Speed Numerical Prediction Error Characteristics And Dynamic Revision Methods To Littoral

As the human society continued to develop, the problems of the global energy shortage and environmental pollution are growing. Wind universality makes it become one of the most suitable new energy for large-scale development and utilization. Accurate forecasting is helpful for the grid interconnection of wind farms and providing safe service. Also, it could reduce the need for transfer capacity requirements and optimize the grid dispatch agenda.The coastal areas have rich wind energy. The wind speed forecasting program has its own peculiarities for discontinuity coastal land surface, wind turbine height being in the atmospheric boundary layer and the difference boundary layer structure between coastal land surface and plain continent. In this paper, wind data observed from Fujian and Jiangsu offshore surface wind towers which is along the coast, MM5 (Mesoscale Model 5) numerical weather prediction data released by the Central Meteorological Observatory and numerical weather data operated by mesoscale numerical model WRF (Weather Research and Forecast) mode are used. Statistical characteristics of numerical model forecast error of wind speed are discussed. According to that, we try to decrease wind speed forecast error and improve forecast accuracy to meet the forecasting techniques of offshore wind farm.Firstly, on the basis of numerical MM5 wind speed forecast error analysis, the dynamic-statistic forecast method of wind speed based on prolonged error is built to predict a year of ultra-short-term wind speed on the wind tower which is around Fujian coastal areas that is on the basis of numerical MM5 wind speed forecast error analysis. And the advantages and disadvantages of that method is analysis by comparing with statistical forecasting methods used in the wind farm. Then, a short-term speed prediction method is proposed by harmonic analysis based on extension forecast effectiveness of the dynamic-statistic forecast method to improve short-term numerical results of wind speed and meet the demand of wind farm. Also, a year data is used to test. Finally, the reasons and characteristics of error caused by numerical models simulation are analyzed by using WRF model results and wind data of Jiangsu coastal wind tower for consideration of the impact of wind speed numerical prediction error caused by coastal discontinuous surface properties and the surface layer-boundary layer parameterization scheme combination. The following conclusions are:The results of a year prediction research on ultra-short-term wind speed of wind tower at Fujian offshore show that the prediction accuracy of wind speed at the wind farm could be improved significantly by using dynamic-statistic forecast method, based on prolonged error, to revise low-resolution numerical model forecasting speed. The mean absolute error (MAE) of wind speed is reduced from 2.45m·s-1,2.50m·s-1 to 1.37m·s-1,1.41m·s-1, respectively. The root mean square error (RMSE) of wind speed is reduced from 3.17m·s-1,3.26m·s-1 to 1.91m·s-1, 1.94m·s-1,respectively. Error index fell about 40%. The line of wind speed that is revised would better reflect the high frequency of wind speed fluctuations and is consistent with the line of the observations. The advantage of that method comparing with single statistical forecast method is that it could forecast dramatic change trend of wind speed when the synoptic system changes greatly or the near ground wind velocity suddenly change to avoid high-capacity impact on the power system for the sudden change.The results of a year prediction research on short-term wind speed of wind tower at Fujian offshore show that short-term (24h) wind speed forecast results could be get with dynamic-statistic forecast method that is based on prolonged error and improved by using harmonic analysis method. The MAE of wind speed is reduced from 2.54m·s-1,2.74m·s-1 to 1.89m·s-1,1.1.95m·s-1, respectively. The RMSE of wind speed is reduced from 3.40m·s-1,3.23m·s-1 to 2.72m/s,2.69m/s, respectively. Those two methods are convenient and economy that could meet the demands of wind farms and be used widely in small and medium-sized wind farms.Choosing different WRF model near-surface layer and boundary layer parameterization combination schemes, the effect of simulation is different. But the simulation capacity of wind speed in autumn and winter is better than that in spring and summer for all the parameterization combination schemes. The simulation capacity of wind speed from land is different from sea. The simulation error of wind speed at hub height is characterized by directivity in offshore wind farm. One of the important reasons is the difference between different underlying surfaces that caused by coastal area terrain characteristics.