An efficient evolutionary optimizer for solving complex dairy feed optimization problems

Abstract

Dairy feed ration optimization is one of the most debated problems of the present time. The presence of linear and non-linear constraints, along with their various potential blends, turns the problem into Non-deterministic Polynomial (NP) hard. This triggers the need for efficient evolutionary algorithms to figure out feed mix at minimum cost and maximum shelf life for dairy cattle. Despite its exclusive importance in a real-life scenario, few papers have developed algorithms to solve this problem. Therefore, an attempt is made in this study to introduce a Self-adaptive multi-population approach with the recently proposed Quadratic Approximation based Jaya (JaQA) algorithm (Das et al., 2022), where care has been taken for the maintenance of diversity in each of its sub-population along with enhancement of convergence. However, considering the difficulty in setting the size of the sub-population, the Self-adaptive Multi-Population Quadratic Approximation guided Jaya (SMP-JaQA) follows an automated learning approach based on the past data to step in the next generation, hence making it a user-friendly approach. The performance of this approach is tested on problems in the benchmark set of Congress on Evolutionary Computation (CEC) 2006 and three popular engineering design functions. Numerical results, statistical tests (including Friedman Rank Test, Iman-Davenport Test, and Wilcoxon signed-rank test), and the convergence analysis confirm the better efficacy of the proposed technique over some recent algorithms. Later, this approach is incorporated into two real-life case studies to optimize the cost of feedstuffs for dairy cattle. The improved results under the feasible domain ascertain the proposed method to be farmer-friendly.