Short-term Wind Speed And Power Forecasting In Wind Farm Based On ANN Combination Forecasting

Energy and environmental concerns are urgent problems of survival anddevelopment of human. Wind power as a kind of renewable and clean energy provides anoption for solving the problem. But wind power is a random and intermittent energysource. Larging scale of wind power connected to the grid, it has negative effects onpower quality, secure and stable operation of power system. Wind speed and powerforecasting in wind farm are effective approachs for the above problems. This paperresearches on wind speed forecasting and wind power forecasting method in wind farm,main work of this paper as follows:①In the base of simulation and calculation of three single forecasting methods, thatis back propagation neural network, radial basis function neural network and supportvector machine, a short-term wind speed nonlinear combination forecasting model baseon general regression neural networks(GRNN) is proposed. The example analysis showsthat generalized regression neural network for nonlinear combination with the predictionresults of three single forecasting methods, can gain higher precision compared to linearcombination method such as single forecasting methods, right average method andcovariance optimization method.②In accordance with the non-linear and non-stability of primitive time series aboutwind speed, taking advantage of empirical mode decomposition theory processing thenonstationary, a short-term wind speed prediction model is proposed based on thecombination of empirical mode decomposition and neural network. The model usesempirical mode decomposition method for preprocessing wind speed at first, whichmakes wind speed into a series of different frequency sequences of the relative steadyweight. Then, according to the different components of the changing rule, properprediction model is established respectively. The high frequency components areforecasted based on generalized regression neural network nonlinear combinationforecasting method, the low frequency components are forecasted based on the RBFneural network forecasting model. Each component of the prediction result is compositedto build the ultimate predictive value. The validity of the proposed method isdemonstrated through two examples, the model obviously increases the predictionprecision.③This paper uses the short-term wind speed prediction model which is based on EMD (Empirical Mode Decomposition) and neural network to forecast the wind speed.From the angle of wind farm，the forecasting of wind farm output power takes the impactof wake effect and different wind directions in consideration when the wind farms arebuilt in different places. This paper derives the formula for calculating the wind farmpower prediction model in all kinds of conditions, and uses the practical examples toprove the validity of the presented models.