基于ASD-KDE算法的超短期风电出力区间预测

Ultra-short term wind power output interval forecast model based on ASD-KDE algorithm

为提高含风电场电网经济调度能力并降低电力系统规划决策的保守性,提出了基于原子稀疏分解-核密度(atom sparse decomposition-kernel density estimation, ASD-KDE)算法的超短期风电出力区间预测模型。该模型应用ASD计算出较为精确的点预测值,并采用粒子群优化正交匹配追踪算法提高原子分解过程的预测实时性。同时针对风电序列不同区域所具有的线性及非平稳特性,构建了衰减线性原子库及Gabor原子库,以期达到自适应分解的效果。再通过对原子分量和残余分量分别进行自预测和BP(back propagation)神经网络预测,获得点预测值。在此基础上,通过对历史风电数据不同区间的划分,构建一维核密度估计模型,逐步滚动获取预测值的置信区间,从而降低了环境变化对预测结果的影响。实际风电场算例验证了所提方法的自适应性、快速性及有效性。

In order to improve the economic operation ability of wind power system and reduce the conservatism of planning and decision making for power system, the ultra - short term wind power output interval forecast model is proposed based on atomic sparse decomposition and kernel density estimation （ASD -KDE）algorithm. ASD is applied to calculate the more accurate point prediction values,and particle swarm optimization algorithm and orthogonal matching pursuit algorithm are combined to improve the real - time performance of atomic sparse decomposition process. Considering the linear and non - stationary characteristics of wind power sequence in different regions, the damped liner dictionary and Gabor dictionary are constructed to reach the self adaptive decomposition, The atomic component and the residual component are separately made self prediction and back propagation （BP）neural network prediction, then the point prediction values can be obtained. And based on that, one - dimensional kernel density estimation model is constructed according to the division of different intervals of wind power data, then the confidence interval for the predicted value can be gained step by step, thus the impact of environmental changes on the prediction results can be reduced. The algorithm example verifies the adaptivity, rapidity and effectiveness of this method for the actual wind power farm.