Establishing the energy consumption prediction model of aluminum electrolysis process by genetically optimizing wavelet neural network

Nowadays, it is very popular to employ genetic algorithm (GA) and its improved strategies optimize neural networks (i.e., WNN) solve the modeling problems of aluminum electrolysis manufacturing system (AEMS). However, traditional GA only focuses on restraining infinite growth optimal species without reducing similarity among remaining excellent individuals when using exclusion operator. Additionally, performing arithmetic crossover or Cauchy mutation, a functional operator that conforms law evolution not constructed generate proportional coefficients, which seriously restricted exploitation hidden potential in algorithms. To above problems, this paper adopts three new methods explore performance enhancement algorithms (EGA). First, mean Hamming distance (H-Mean) metric designed measure spatial dispersion alleviate selection pressure. Second, with transformation sigmoid-based function developed dynamically adjust exchange proportion offspring. Third, an adaptive scale coefficient introduced into Gauss-Cauchy can regulate mutation step size real time search accuracy for population. Finally, EGA solver employed deeply mine initial parameters wavelet network (EGAWNN). Moreover, provides test, convergence analysis significance test. The experimental results reveal EGAWNN model outperforms other relevant wavelet-based forecasting models, where RMSE test sets based 305.72 smaller than seven