Short-term Wind Power Forecast Based On Dynamic Model

With the rapid growth of wind power industry in China, the installed capacity is increasing and the proportion of wind power grid in our country is becoming bigger, However, because of the volatility and intermittent of wind power which impacted by uncertainty and instability of wind, large-scale grid-connected wind power will inevitably bring serous problem to the operation of power system. While an accurate power prediction can improve the stability of power system, therefore, to study how to improve the dynamics and accuracy of power prediction have an important significance on the dispatching and security of power system. This paper puts forward a modified wind power intelligent forecast model by analyzing the advantages and disadvantages of existing wind power prediction method. The specific research content is as follows.First, to combine with the characteristics of wind illustrates the factors which involved in wind power, and according to the laws of the wind data of wind speed and direction, the prediction model is established for the determination of input vector provide guidance.Second, in order to strengthen the dynamic performance of Elman network, the structure and incentive function was modified by studying the neural network. On the other hand, local mean decomposition method is used to input data decomposition and LMD-Elman short-term dynamic power forecasting model is established. Experiments shows that this method enhances the dynamic of wind power prediction.Third, As to the problem of falling into the local optimum in standard artificial bee colony, it is proposed to introduce an adaptive factor which can expand the search of the swarm and use the Cauchy distribution to improve the universality of colony search, This improved algorithm named adaptive Cauchy mutation artificial bee colony(ACMABC). Then, using the ACMABC algorithm optimized the weight parameters of the NARX neural network and Elman neural network model, a wind power forecasting model based on the dynamic power forecasting model is presented for forecasting, comparison with the NARX neural network model and Elman neural network model. The simulation and analysis prove that the new model has a higher precision.Finally, on the basis of above research, a kind of method by using Labview and MATLAB mixed programming to realize the control and prediction of system. This method not only reduces the design of the program code, but also makes the programming graphical visualization and reduce the difficulty.