Study On The Model Of Short-time Wind Farm Generated Power Forecasting

As fossil fuel consumption is unceasing and environment problems are becoming serious in the Earth, wind energy, as a safe, clean and limitless energy, provides the opportunity to change the global energy mix. Wind power is the main form of wind energy, and many countries make the development of wind power as national development planning. Due to the random and uncontrollable nature of the natural wind, the output of the wind farm power fluctuations. While wind farm penetration is over a certain value, it will effect power quality and security, stability operation of grid network system. Accurate wind power forecasting can change scheduling plan, system reserve capacity and other arrangements to reduce the overall active power output fluctuations of wind power, thus to improve penetration power, with reducing the impact to the power system and improving the competition ability in power market of wind farm.Accuracy of prediction will be the impact of forecasting model selection, and the selection of forecasting model is based on the suitable mathematical methods, choosed by the properties of data series. This paper reconstructs the phase space for the wind generated power time series by using the advantages of high reliability, fast computation of C-C algorithm which can calculate the delay time and embedding dimension meantime, and proves that the wind generated power time series has the chaotic characteristics through calculating the correlation dimension and the largest Lyapunov index, so chaos theory can be used for wind power short-time forecasting.Based on the phase space reconstruction, different order of Volterra adaptive filters is used to forecast short-time wind generated power. Analysis results show that the results forecasted by the designed forecasting model can reflect future changes trend of wind generated power and possess high precision, while different order of Volterra adaptive filters has different accuracy, the lower the order is, the higher accuracy is. In addition, the RBF neural network model is used to predicte wind power, and the RBF neural network model can achieve higher prediction accuracy for the wind farm power generation, while there is certain lag characteristics. The two motheds above only need wind power time series and have simple realization, high precision, so it can provide foundation for wind power chaotic prediction.Similar days has been used for power load prediction and achieved good result. According to the factors impacting wind power output, a method to select similar days of wind farm's generated power is proposed and a short-term wind power forecasting model based on similar days and Chebyshev neural network is designed:wind power output contained in a large number of history days and their similar days is used to train Chebyshev neural network, and put the data of similar days of the the forecasted day to the trained Chebyshev neural network and power output prediction is gotten. The designed model is verified by the data of a certain wind farm located in Yunnan province, compared with continuous method, and the forecasting error and its probability distribution are analyzed. Analysis results show that the results forecasted by the designed forecasting model possess high accuracy and forecasting error meets normal distribution, so it is available for reference in wind power forecasting.