A grey theory based back propagation neural network model for forecasting urban water consumption

Forecasting urban water consumption is a complicated task due to its unavoidable huge fluctuation caused by uncertain factors. Back propagation neural network (BPNN) is known for its strong ability to deal with nonlinear problems but is limited by the requirement for large samples and relatively high computation complexity, while grey theory has advantages such as requiring little samples and easy modeling and computing. Therefore, a combined grey theory and BPNN model named GM-BPNN is proposed is proposed to forecast urban water consumption in Hangzhou. Simulation results show that GM-BPNN can reduce the value of mean absolute percentage error (MAPE) by 6.25% and 4.62% compared with GM (1,1) and original BPNN which means GM-BPNN achieves higher prediction accuracy.