基于ICA-NN的短期风功率预测研究

Research on Short-term Wind Power Prediction Based on ICA-NN

随着风电大规模的接入电网,风电对电网的影响越来越大。由于风电出力具有随机性、间歇性和不可控性,导致风电对电网调度运行带来巨大的挑战。为了充分利用风电,必须将风电由未知变为基本已知,提高对风电出力的预测精度。提出一种基于帝国主义竞争算法的神经网络(ICA-NN)方法来提高短期风功率预测的精度。在该方法中,首先,建立一个基于多层感知器(MLP)人工神经网络的风速预测模型,然后,用帝国主义竞争算法优化神经网络中的权值。将该预测方法应用于新疆某风电场,验证了该方法应用于短期风功率预测的有效性,证明了该方法可以提高短期风功率预测的精度。

With the large - scale access of wind power grid, the wind power has more and more influence on power grid. Be- cause the wind power output is random, intermittent and uncontrolled, it brings the huge challenge to power grid dispatching and operation. In order to make full use of wind power, the wind power must be changed from unknown to known, and the prediction accuracy of wind power output should be improved. A neural network method based on imperialist competitive algo- rithm （ ICA - NN） is presented to improve the accuracy of short - term wind power prediction. In this method, first of all, a prediction model of wind speed is established based on artificial neural network with multi - layer perceptron （ MLP）, and then, the weight values of neural network are optimized with the imperialist competitive algorithm. The prediction method has been applied to a wind farm in Xinjiang, which verifies its effectiveness in short - term wind power prediction, and proves that the proposed method can improve the accuracy of short - term wind power prediction.