基于BAS-BP模型的光伏并网短期负荷预测方法

A Short-Term Load Forecasting Method for Photovoltaic Grid Connection Based on BAS-BP Model

中文关键词:负荷预测  短期负荷  BAS-BP模型  光伏发电

keywords:load forecasting  short-term load  BAS-BP model  photovoltaic power generation

中文摘要:针对光伏并网场景下短期负荷预测精度不足的问题，提出一种基于天牛须搜索算法-反向传播（beetle antennae search-back propagation，BAS-BP）模型的光伏并网短期负荷预测方法。首先，依据不同的电力能源用户，分析光伏并网负荷特性；其次，通过补全缺失数据、替换异常数据以及归一化数据量纲，完成历史负荷数据的预处理；最后，通过天牛须优化算法来优化反向传播神经网络，以预处理后的气象、负荷、时间等特征数据为输入，构建BAS-BP光伏并网短期负荷预测模型，据此获得短期负荷预测结果。实验结果表明：所提方法预测结果的平均绝对误差、平均相对误差、均方误差的最小值分别为0.2、0.16、0.3，验证了该方法的应用效果良好。

Abstract:Aiming at the problem of insufficient accuracy in short-term load forecasting under photovoltaic grid-connected scenarios， a short-term load forecasting method for PV grid connection based on the beetle antennae search-back propagation （BAS-BP） model is proposed. Firstly， the load characteristics of PV grid connection are analyzed according to different power users. Secondly， historical load data are preprocessed by filling missing data， replacing abnormal data and normalizing data magnitudes. Finally， the back propagation （BP） neural network is optimized by the beetle antennae search （BAS） algorithm. Taking the preprocessed meteorological， load and time characteristic data as inputs， the BAS-BP short-term load forecasting model for PV grid connection is constructed to obtain short-term load forecasting results. Experimental results show that the minimum values of mean absolute error， mean relative error and mean square error of the proposed method are 0.2， 0.16 and 0.3 respectively， which verifies the good application effect of the method.