Abstract: Protected agriculture plays a crucial role in enhancing agricultural product supply and resource-use efficiency in the context of climate change. Among various crops cultivated under protected conditions, tomato stands out as one of the most representative due to its high economic value and widespread production. Its growth rate and plant health status exhibit significant sensitivity to environmental conditions, making precise environmental management essential. Given the complex and resource-intensive nature of tomato production, it is often necessary to regulate multiple environmental factors within greenhouses to optimize yield and quality. However, existing research predominantly focuses on optimizing static or quasi-static performance indicators, lacking sufficient characterization of the coupling relationships among multiple environmental factors and their dynamic evolutionary characteristics in greenhouse environments. Consequently, the stability and adaptability of optimization results remain inadequate when confronted with real-world disturbance conditions, such as sudden climatic shifts or equipment fluctuations. To address these limitations, this study proposes a stability-aware multi-objective optimization framework, the Chaos-Lyapunov-enhanced Non-dominated Sorting Genetic Algorithm II (CL-NSGA-II). Such integration enables more robust optimization in complex agricultural ecosystems with inherent nonlinear dynamics and time-varying interactions. Building upon the traditional NSGA-II algorithm, this method introduces chaotic mapping to enhance adaptive perturbation capability during population initialization and the search process, thereby improving global exploration and avoiding premature convergence. Furthermore, it incorporates Lyapunov exponents to evaluate the stability characteristics of the system during dynamic evolution, ensuring that solutions exhibit long-term operational stability. This dual mechanism achieves a dynamic balance between exploration ability and convergence performance, allowing for the identification of solutions that are both high-performing and dynamically reliable. On this basis, this paper constructs a multi-objective optimization simulation model for the greenhouse tomato cultivation process, applying the proposed CL-NSGA-II algorithm to the coordinated optimization of key environmental factors, including temperature, humidity, light intensity, and CO₂ concentration. Firstly, growth rate and plant health data were collected under various combinations of temperature (20～35℃), humidity (50%～80%), and light intensity (400～700μmol/(m²·s)) in a controlled environmental chamber, forming a comprehensive multi-objective optimization dataset. Secondly, chaotic mapping was embedded to initialize the population, with adaptive dynamic perturbation enhancing search diversity. Simultaneously, a Lyapunov exponent constraint was applied during evolution to solutions with negative exponents (λ<0), thereby ensuring stability in the resulting environmental control strategies. These methodological innovations collectively address the critical challenge of maintaining operational stability while pursuing optimal crop performance.Experimental results indicate that, compared to the standard NSGA-II algorithm, the proposed CL-NSGA-II algorithm improves solution set coverage by 19.7%, accelerates convergence speed by 15%, and reduces the Lyapunov exponent to -0.054, demonstrating significantly enhanced system stability. This approach not only effectively coordinates the growth rate and plant health of greenhouse tomatoes but also ensures the long-term stable operation of the environmental system. It thus provides methodological support for developing intelligent control strategies and optimizing parameters for tomato cultivation in protected agriculture, contributing to more efficient, sustainable, and climate-resilient agricultural production systems.