Research On Short-term Prediction Methods Of Wind Speedand Wind Powerat Wind Farm

Wind power generation technology is a renewable energy which is becoming more mature,lower cost and rapid developing. Now our country has become the fourth largest Wind powergeneration country which has wind power installed capacity more than kilowatts of wind powergeneration power. The influence of wind power is more and more extensive. Wind power hasvolatility, low energy density, intermittent characteristics, which causes wind power outputpower also has the same characteristics. High randomness of wind brings a big problem for awind farm to provide electric grids with necessary power when the wind power penetrationpower more than8%. It seriously influences power system, safety set operation and powerquality. So the wind power short-term forecast is of great significance for reasonablearrangement of dispatching plan. To improve the ability of power system acceptance windpower.This paper is based on the raw data of in Inner Mongolia wind farm，and the short termprediction model of wind speed and wind power is researched. Steady history wind power datacan be regarded as the typical non-stationary random process. Firstly, ARMA time seriestechnique is researched in detail. Use the finite difference method for non-stationary time seriesinto stationary time series. After analysis the model of ARMA (2,3) is developed and used thismodel to forecast the wind speed. Then calculate the max Lyapunov exponent of wind speed andwind power, both of their Lyapunov exponents are greater than zero. It indicting the wind speedand wind power have the characteristics of chaotic characteristics. So this paper use RBF chaoticRBF neural network model to forecast. In order to improve the prediction accuracy, using C-Cmethod to conduct joint reconstruction parameters optimization.Considering any a model has its own advantages and disadvantages, in order to disperse therisk of prediction, finally decide use the combined model of the minimum predict error sum ofsquares to forecast wind speed. This paper got the wind prediction power according to theforecasted wind speed and wind power curve of wind turbines. This is called the power curvetransformation method. Because wind power has the characteristics of chaotic characteristics, sothen use RBF chaotic RBF neural network model to forecast wind power directly which is calledthe direct forecasting method.By comparison of all prediction results, we find out that wind speed forecasting precision ishigher than wind power prediction. It shows that wind speed has more regularity. Predictionaccuracy of direct forecast method is higher than the power curve transformation method. So it is suggested that use direct prediction method when based on chaos theory of wind powerprediction.