基于多重降噪的改进SSA-LSSVM短期电力负荷预测模型

Improved SSA-LSSVM Short-term Power Load Forecasting Model Based on Multiple Noise Reduction

针对电力负荷数据的波动性,提出了一种基于多重降噪和改进的麻雀搜索算法(improved sparrow search algorithm,ISSA)优化的最小二乘向量机(least squares support vector machines,LSSVM)预测模型。首先,采用自适应小波阈值降噪对原始数据进行降噪处理;然后,采用完全自适应噪声集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)方法分解数据,再将分解后较复杂的分量进行奇异值分解(singular value decomposition,SVD)降噪处理;最后,采用多策略改进麻雀搜索算法对LSSVM方法的参数进行优化,对分解后的数据进行预测叠加。这种组合模型实现了对数据的多重降噪和对ISSA算法的优化,能够有效提高预测精度。与SSA-LSSVM、ISSA-LSSVM、CEEMDAN-ISSA-LSSVM模型相比,所提出的组合模型平均绝对百分比误差分别降低52.24%,25.58%,15.79%。该结果表明,所提组合模型能够有效预测短期电力负荷。

Aiming at the volatility of power load data,an optimized least squares vector support machine(LSSVM) prediction model based on multiple noise reduction and improved sparrow search algorithm(ISSA) is proposed.First,adaptive wavelet threshold denoising is used to denoise the original data,and then the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) method is used to decompose the data,and then singular value decomposition(SVD) is used to denoise the more complex components after decomposition.Finally,the parameters of LSSVM are optimized by using multi strategy improved sparrow search algorithm,and the decomposed data are predicted and stacked.This combined model can effectively improve the prediction accuracy after multiple noise reduction of data and ISSA algorithm optimization.Compared with SSA-LSSVM,ISSA-LSSVM and CEEMDANISSA-LSSVM models,the average absolute percentage errors of the proposed combined model are reduced by 52.24%,25.58% and 15.79% respectively.The results show that the proposed combined model can effectively predict short-term power load.