Optimization For Short-term Wind Power Prediction Method

Short-term wind power prediction has great significance for scheduling and operation of electric power system as well as production and maintenance of wind farm. The foreign research of short-term wind power prediction began in 1980s. With development of research, the quite mature algorithms and commercial softwares have been applied. But in China, the related studis just developed in recent years and there were still many shortage in the forecast precision, the reliability and the adaptability of different wind farm. Therefore it has good academic value and practical value to study how to improve forecasting accuracy and reliability of the algorithm for wind power prediction based on numerical weather prediction.In the paper, the following aspects of work was conducted:Firstly, statistical analysis of the numerical weather prediction error and wind power short-term prediction error and the error distribution are studied and the influnce of invalid data in the measured data, numerical weather prediction error and prediction models are discussed for short-term wind power prediction.Aimed at the fault of existing interpolation method, the fuzzy inference system is the first time introduced into the interpolation of missing wind wind speed and direction data. Historical data series of measured wind speed and direction are used to add the missing data to reduce the error. Through this method, the invalid data in the measured data can be corrected, so as to lay a good foundation for correcting numerical weather prediction data.Numerical weather prediction is the main error source of short-term wind power prediction and the current linear correction methods can not correct the error very well. A nonlinear correcting algorithm based on the wavelet packet analysis algorithm and BP neural network is proposed to correct the numerical weather prediction output. Then the corrected data are used to predict the wind power output. Fanally forecasting output power are corrected angin to get the final forecasting output power. Through the data correction algorithm, the error of numerical weather prediction are reduced and the accuracy of short-term wind power prediction are improved.A improved particle swarm optimization algorithm are used to optimize the existing short-term wind power prediction moedl with BP neural network. The global search capability of particle swarm optimization algorithm and the local search capabilities of BP neural network are combined to improve the power output of wind power prediction model.Finally the uncertiaty and the probability density function of short-term wind power prediction are discussed. The confidence and confidence intervals are used to assess the uncertianty of short-term wind power prediction.