基于BA-VMD-LSSVM的超短期电力负荷预测

Ultra-short-term Load Forecasting Based on BA-VMD-LSSVM

超短期电力负荷具有随机性强、波动性大等特点,使得对其进行高精度的预测比较困难。文中提出基于全局参数优化的超短期负荷预测模型。在训练阶段,建立平均绝对百分比误差(MAPE)作为蝙蝠算法(BA)的目标函数,以优化变分模态分解(VMD)、最小二乘支持向量机(LSSVM)及输入数据点数。在测试阶段,应用设置最优参数的VMD分解负荷数据,并使用LSSVM处理各分量,以完成对电力负荷的高精度预测。数据分析结果表明,使用BA对VMD、LSSVM和输入数据点数进行全局优化能够有效地提高超短期电力负荷的预测精度。

Ultra-short-term power load possesses such features as strong randomness and volatility,which makes it difficult to predict load data.Ultra-short-term load forecasting based on global optimization is proposed.During the training stage,mean absolute percentage error(MAPE)is used as target function of bat algorithm(BA)to find the optimal parameter of variational mode decomposition(VMD),least square support vector machine(LSSVM)and the entered number of data points.During the testing stage,the optimized VMD is applied to decompose load data,and then LSSVM is employed to analyze each component.Data analysis indicates that the optimization of VMD,LSSVM and the entered number of data points using BA can improve prediction accuracy.