<p>The RIME algorithm is a novel physical-based meta-heuristic algorithm with a strong ability to solve global optimization problems and address challenges in engineering applications. It implements exploration and exploitation behaviors by constructing a rime-ice growth process. However, RIME has some drawbacks such as the tendency to fall into local optima and the imbalance between the exploration and exploitation phases. In order to overcome the limitations of the RIME algorithm, this paper introduces an Augmented Multi-Strategy Improved RIME Algorithm (AMRIME). Firstly, the elite pool soft-rime search strategy based on Levy and Brownian motion is designed to help the algorithm achieve a better balance between exploration and exploitation. Next, a leader hard-rime puncture mechanism based on covariance learning is proposed in this paper, which enables the algorithm to select the search direction better by absorbing the location information of excellent populations during the search process. To deal with the deficiency of premature convergence of RIME and to enhance its local escaping ability, this paper proposes a local escaping strategy. To solve the problem of population scarcity in iteration, a quasi-reverse learning strategy is introduced to enhance population diversity and ensure convergence accuracy. The proposed AMRIME is compared with other basic or improved algorithms using the CEC2017, CEC2019 and CEC2022 test sets, and moreover, the effectiveness of the proposed strategy is evaluated. By analyzing the convergence curves, box plots, Wilcoxon rank sum test and Friedman test, it is demonstratede that the RIME algorithm exhibits a stronger ability to evade local optima and achieves higher convergence accuracy compared to other metaheuristic algorithms. Finally, the AMRIME algorithm is applied to solve the load distribution of chillers. The results show that AMRIME is able to obtain the optimal load distribution to reduce energy consumption compared to other comparative algorithms.</p>