SKBA-LSSVM短期光伏发电功率预测模型

SKBA-LSSVM Short-term Forecasting Model for PV Power Generation

为了提高短期光伏发电的预测精度,减少光伏发电不稳定性对于用户和电网的影响,提出一种结合相似日理论和K-means改进蝙蝠算法优化最小二乘法支持向量机LSSVM(least squares support vector machine)正则化参数和核参数的光伏发电功率短期预测方法。该方法通过历史数据集和预测日数据分析影响光伏发电功率的因素,构建日特征向量,筛选历史日数据作为训练集,并将预测日数据作为校验集。利用改进的蝙蝠算法全局寻优特性对LSSVM的参数进行优化,构建短期光伏发电功率预测模型。将所提模型与其他智能算法进行比较,结果表明该方法预测精度较高。

To improve the prediction accuracy of short-term photovoltaic(PV)power generation and reduce the impact of the instability in PV power generation on users and power grid,a short-term forecasting method for PV power genera tion based on similarity day theory and K-means improved bat algorithm(KBA)is proposed to optimize the regulariza tion and kernel parameters of least square support vector machine(LSSVM).This method analyzes the factors affecting the PV power generation by using historical data sets and data on forecasting dates,constructs daily feature vectors,fil ters the data on historical dates as a training set,and uses the data on forecasting dates as a test set.The parameters of LSSVM are optimized by using the global optimization characteristics of the improved bat algorithm,thus a short-term forecasting model of PV power generation is constructed.From the comparison between the proposed model and other in telligent algorithms,it is shown that this method has higher prediction accuracy.