Research On Short-term Wind Power Forecasting Of Grid-connected Wind Farm

Wind is the energy with most developed renewable potential for its clean environment, mature technology and low develop cost. But the characteristics of intermittent, randomness and anti-regulation of wind power make grid connecting wind power generate many adverse effects on the grid scheduling and running. One major way to overcome the difficulties is the wind power prediction and control technology. In this paper, the wind farm power forecast is achieved based on the measured data of a wind farm in Zhang jiakou.(1)Because of the characteristics of randomness and volatility of the wind speed, there may be a large amplitude variation in a short time. Wavelet function with certain characteristics(db5) is chosen to take advantage of the multi-resolution characteristics of wavelet decomposition. Low-frequency approximation with a similar trend and the high frequency detail components containing the noise are obtained after the sequence of the original wind speed is decomposed into a certain number of layers.(2)The wind speed data acquired in wind farm has a strong instability. In this paper, the prediction on base of genetic algorithm and least squares support vector machine is proposed taking advantage of the genetic algorithm in solving complex problems. The genetic algorithm optimization least squares support vector machine models of each sequence of the wavelet decomposition are established.The wind speed prediction is acquired based on the recombination of prediction outcome.(3)It is necessary to establish the power curve in line with the wind farm characteristics to improve the prediction accuracy of the power. Because it has a strong non-linear relationship between wind speed and power, the genetic algorithm optimization RBF neural network wind- power conversion model is established based on the advantages of RBF neural network in curve fitting. A mature network model is acquired based on the original speed and power as training data. The prediction results are obtained based on the forecast wind speed data as the test data.To verify the validity of the model, two different wind speeds data are selected, and the results show that the model can predict with good results.