ABSTRACT
Distributed Generation (DG) is an intermittent power supply, environmental factors change has great influence on its power output, so the output power of the distributed generation is inconclusive. It is not beneficial for the power grid dispatchers to dispatch the conventional power supply with the distributed generation. Therefore, we need to forecast output power of the DG and obtain the power output development curve, then, coordinated controlling the conventional power supply with the DG to reduce its influence on power system and improve security and stability of the operation of power system. There are two main methods used for DG output forecasting: indirect method and direct method. The indirect method, is through the meteorological station or the information of the weather forecast, according to the change of relative parameters of environment. Through the calculation of a certain mathematical model to obtain the prediction of the DG output. There are commonly used methods, such as neural network algorithm. Due to the sun light characteristics, S.-X. Wang and N. Zhang proposed a combination-forecasting model based on the gray neural network in the literature. Atsushi Yona, Tomonobu Senjyu and Ahmed Yousuf Saber used three different methods of artificial neural networks: feedforward neural network, radial basis function neural network and recurrent neural network, to predict distributed
relationship between the data, and building prediction model. The phase space reconstruction is also called dynamic system reconstruction. Through the one-dimensional time series, reverse constructed the phase space structure of original system. Delay vector method is a commonly used method, its basic idea is the evolution of the arbitrary component in the system is determined by other components which interact with it, thus, the information of these relevant components are implying in the process of development of arbitrary component.
