基于ISSA-BPNN模型的城镇燃气日负荷预测  

Urban Gas Daily Load Prediction Based on ISSA-BPNN Model

作　　者：肖荣鸽[1] 夏海平 李雨泽 XIAO Rongge;XIA Haiping;LI Yuze

机构地区：[1]西安石油大学石油工程学院,陕西西安710065

出　　处：《煤气与热力》2025年第2期I0016-I0022,共7页Gas & Heat

摘　　要：针对单一神经网络模型在预测过程中存在精度低、易陷入局部最优等问题,提出了一种改进的麻雀搜索算法优化的BP神经网络(ISSA-BPNN)模型用于燃气日负荷预测。BP神经网络与优化算法结合的模型预测精度高于单一BP神经网络。选择、交叉、变异等操作可以提高种群多样性,避免算法过早收敛。ISSA-BPNN模型平均绝对百分比误差为0.0393,在组合模型中预测精度最高,耗时也最长。An improved sparrow search algorithm to op-timize BP neural network(ISSA-BPNN)model is pro-posed for gas daily load prediction to address the is-sues of low accuracy and susceptibility to local optima in the prediction process of a single neural network model.The prediction accuracy of the model combin-ing BP neural network and optimization algorithm is higher than that of a single BP neural network.Opera-tions such as selection,crossover and mutation can im-prove population diversity and avoid premature conver-gence of algorithms.The average absolute percentage error of ISSA-BPNN model is 0.0393,which has the highest prediction accuracy and longest time consump-tion in the combined model.

关 键 词：燃气日负荷 BP神经网络 改进的麻雀搜索算法 负荷预测 

分 类 号：TU996.6[建筑科学—供热、供燃气、通风及空调工程]