基于混核LSSVM的批特征风功率预测方法

Wind Power Prediction Method Using Hybrid Kernel LSSVM With Batch Feature

针对风电场风功率预测问题,利用历史风功率、气象数据和测风塔实时数据等相关信息,提出了带有批特征的混核最小二乘支持向量机(Hybrid kernel least squares support vector machine,HKLSSVM)方法,建立风电场风功率预测模型.为了增强模型的适应性,设计改进的差分进化算法对模型参数进行优化,并利用稀疏选择方法来选取合适的训练样本集,缩短建模时间,保证预测模型精度.根据风场风机的地理位置分布情况,提出批划分的建模策略,对相近地理位置的风机进行组批,替代传统风场风功率预测方法.通过风场中实际数据进行测试,实验结果表明与其他预测方法相比,本文提出的方法能够提高预测精度和效率,减少风电波动性对电网的影响,从而提高电网的安全性和可靠性.

For the wind power prediction problem in a wind farm,this paper collects some related data such as historical wind power data,meteorological data,and wind speed data sampled by anemometer tower.Then,a wind power prediction method with batch feature is proposed,which is based on hybrid kernel least squares support vector machine(HKLSSVM).It is used to establish the wind power prediction model in the wind farm.To enhance the model’s adaptability,an improved differential evolution algorithm is designed to optimize the model parameters,and a sparse selection method is used to select the appropriate training samples set.Thus,the modeling time is shortened and the prediction model accuracy is guaranteed.According to the location distribution of wind turbines in the wind farm,a modeling strategy based on batch partition is proposed,some wind turbines at similar locations can be clustered by batch strategy,which is used instead of the traditional wind power prediction methods in the wind farm.The proposed model is tested through the real data in the wind farm.Experimental results show that the proposed method can improve the accuracy and efficiency of wind power prediction compared with other prediction methods,and can reduce the effect of the wind fluctuation.Hence it can ensure the safety and reliability of the power grid.