基于GRA-ABC-BPNN模型的城市燃气日负荷预测

Daily load forecasting of urban gas based on GRA-ABC-BPNN model

城市燃气负荷预测对于合理高效地调配燃气资源、解决城市燃气用户用气问题具有重要意义。通过灰色关联分析法(Grey Relation Analysis,GRA)对所确定的11个影响燃气日负荷的因素进行分析,依据关联度大小进行筛选,逐个剔除关联度较低的影响因素,将剩余关联度较高的影响因素作为BP神经网络(Back Propagation Neural Network,BPNN)的输入;采用人工蜂群(Artificial Bee Colony,ABC)算法优化BPNN权值和阈值;搭建GRA-ABC-BPNN预测模型预测城市燃气日负荷,并对其准确性和有效性进行验证。结果显示:GRA-ABC-BPNN模型预测的城市燃气日负荷的平均绝对百分比误差(Mean Absolute Percentage Error,MAPE)值为0.5528%,而遗传算法优化BPNN(Genetic Algorithm-BPNN,GA-BPNN)模型及ABC-BPNN模型的MAPE值分别为1.4913%、0.6369%,证明了GRA-ABC-BPNN预测模型是一种有效且精度可观的城市燃气日负荷预测方法,为城市燃气日负荷预测提供了新的途径。

Urban gas load forecasting is of great significance for rationally and efficiently deploying gas resources and solving the problem of gas consumption by urban gas users.Herein,the 11 identified influencing factors of the daily gas load were analyzed through the Gray Relation Analysis(GRA)method and screened according to the correlation degree,having the influencing factors with little correlation eliminated one by one,and using the remained influencing factors with high correlation degree as the input of the Back Propagation Neural Network(BPNN).Meanwhile,the BPNN weights and thresholds were optimized with the Artificial Bee Colony(ABC)algorithm.Besides,a GRA-ABC-BPNN forecasting model was established to predict the daily load of urban gas,and the accuracy and effectiveness of the established forecasting model was verified.As shown by the results,the Mean Absolute Percentage Error(MAPE)of the daily load of urban gas forecast by GRA-ABC-BPNN model is 0.5528%,while the MAPEs of Genetic Algorithm-BPNN(GA-BPNN)model and ABC-BPNN model are 1.4913%and 0.6369%,respectively.This indicates that the GRA-ABC-BPNN forecasting model is an effective and accurate method to forecast the daily load of urban gas,and it could provide a new way for daily load forecasting of urban gas.