Research On The Application Of Intelligent Algorithms In The Short-term Wind Power Prediction

In recent years, with the emphasis on renewable energy industry, large-scale wind power is increasingly involved in power grid step by step, and has played a positive role for improving the structure of the power supply. Moreover, as the global energy crisis intensified, the development and application of clean and renewable energy is imminent, and wind energy is a renewable and clean energy, wind power is the most large-scale development of technical and economic conditions of hydropower renewable sources of energy. So the wind power gradually becomes the world recognized to be bestowed favor on newly, which can effectively relieve the energy crisis and promote the sustainable development of low-carbon economy.But wind power has the disadvantages of volatility, intermittent and low energy density. Great wind power fluctuations will bring adverse effects to power balance and frequency regulation of power grid and put a threat on the security, stability, economic and reliable operation of the power system when large-scale wind farms connect to the electricity grid. Therefore, it has already been a currently needed to be analyzed and solved practical problem that accurately predicting the wind power, combining the grid organically with the inherent characteristics of the operation and effectively playing the value of wind power. Taking wind power prediction into further study is very important to improve the ability of wind farm and power system coordination operation and promote the sustained and healthy development of wind power.This paper focuses on the following research mainly centered on the short-term wind power prediction based on the data of history wind power:1. With the characteristics of rich theory and high precision, neural network model is a common nonlinear model which can better cope with the fluctuations of sequence. Wavelet decomposition and reconstruction is applied to process data, and then Wavelet-BP neural network model is built to predict wind power.2. Modeling simple, time series model has good short-term prediction accuracy. Analyzing the principle of time series method and building ARIMA model according to the characteristics of non-stationary power sequence to predict short-term output power.3. With the advantages of fast convergence speed, strong learning ability and good generalization ability, SVM can effectively predict the change trend of sequence. Based on the principle of SVM, the regression model of SVM is built to predict the wind power directly.4. Studying on model parameter optimization problems of support vector machine (SVM). It analyzes the effect of penalty factor and nuclear parameters and the impact of the performance of SVM. the genetic optimization algorithm is applied to propose the optimized support vector machine forecasting model (GA-SVM) based on genetic algorithm to deal with the insufficiency of SVM modeling. Finally, we use GA-SVM prediction model for wind power prediction.5. Studying on wind power combination prediction model which combines the wavelet-BP model, time series model and GA-SVM model. Aiming at the critical problem of combination model of solving the weight coefficient, respectively, using the minimum variance method and optimal non-negative variable weighting coefficient method to gain the weighted coefficient of various forecasting methods, so as to build combination model for power prediction.