<p>Serial-batching scheduling problems, characterized by processor learning and job deterioration in uncertain processing environments, are prevalent in practical production systems. While serial-batching and fuzzy scheduling problems have been extensively explored individually, the integration of both aspects remains underexplored. This study investigates serial-batching scheduling problems that incorporate learning and deterioration effects both on a single machine and on non-identical parallel machines within a fuzzy environment. Given the uncertainty prevalent in practical processing environments, we model the processing time of each job as a trapezoidal fuzzy number. For the single machine scenario, we identify several structural properties and propose a heuristic algorithm, HA1. In the context of non-identical parallel machine scheduling, the inherent NP-hardness of the problem prompts us to introduce another heuristic algorithm, HA2, to generate a preferable initial solution. Subsequently, we develop a hybrid algorithm named IPSO-HA2, which integrates an Improved Particle Swarm Optimization (IPSO), the heuristic algorithm HA2, and a Variable Neighborhood Descent (VND) local search strategy, to find optimal or near-optimal solutions for the parallel machine scheduling problem. Algorithm HA1 is then applied to determine the optimal sequence of batches and jobs on each machine. Computational experiments demonstrate that the proposed heuristic algorithms HA1 and HA2, along with the improvement strategy, significantly enhance the performance of the hybrid algorithm IPSO-HA2. The proposed algorithm outperforms the Biased Random-key Genetic Algorithm (BRKGA), Standard Particle Swarm Optimization (SPSO), Differential Evolution (DE), and Artificial Bee Colony (ABC) in terms of minimizing the maximum completion time.</p>