An enhanced gene expression programming for daily water consumption forecasting with new chromosome structure

Abstract

Accurately forecasting the urban daily water consumption is crucial for planning water supply. However, the daily consumption data obtained is limited, usually only one data per day, which restricts the performance of existing algorithms. To address this challenge, this paper developed an enhanced gene expression programming (MulBE-GEP) model for daily water consumption forecasting, which utilizes an innovative chromosome structure with multiple basal and enhanced genes. During the forecasting process, an improved phase space reconstruction method based on KL divergence analysis was proposed to achieve adaptive data reconstruction. Then, the optimal forecasting model and its mathematical expression were obtained through the MulBE-GEP. In addition, an innovative fitness function Base Score (BScore) is designed to improve forecasting performance. Experiments show that the proposed forecasting model can improve performance by 1.2% ~ 8% under multiple metrics and fully demonstrate the fast and stable optimization ability compared to the other seven algorithms.