| With the decreasing conventional fossil energy in the world and global warming, manenhances the development and utilization of new energy. Wind energy, which is a kind ofclean, abundant and renewable energy, is different from fossil fuels in nature. Owing to thewind’s volatility, randomness and intermittence, the large-scale connection of wind farms withpower grid has adverse effect on the safety, economic operation and power quality of powersystem. Accurate wind power prediction will reduce the number of grid spare capacity and theoperation cost of power system. Also it is very important to participate in the electricitymarket and reasonable scheduling of power grid. In this paper, measured wind power data, ata farm in Shandong Province, is used to forecast short-term wind power.First of all, the analysis indicates that the main factors affecting wind power are windspeed, wind direction, temperature, pressure and relative humidity. Based on these factors, theBP and RBF model are established and the influence degree of the prediction model isanalyzed by the latter three factors.Then, chaos theory is applied. It’s proved that wind power sequence has chaotic character.In order to get the potential rule, phase space reconstruction is applied and C-C method isused to get the best parameters of the reconstructed phase space. Meanwhile, the weightedone-order local model is used to predict wind power.Finally, the improved wavelet-BP neural network method is combined with the globaloptimization search ability of genetic algorithm, the good time-frequency local property ofwavelet analysis and BP neural network’s powerful nonlinear mapping ability. It uses thechaos phase space reconstruction theory to determine the BP model’s node number in theinput layer and has a very high prediction precision. At the same time, this improved methodis carried on the wind speed prediction, and the prediction power results are obtained with thetransformation of the actual power curve. The results show that in regularity, wind speed isbetter than wind power. With the conversion of the measured power curve, the powerprediction accuracy goes bad. |
