This paper proposes an opportunity maintenance decision model based on variable time windows for serial-parallel systems with buffer in batch production environments, aiming to enhance system operating efficiency and reduce downtime costs caused by equipment failures. First, considering the dynamic variation of equipment failure rates across batches, an estimation model for the average work-in-progress (WIP) quantity in buffers under different batches is established using the decomposition method, reflecting the differentiated buffering effects of buffers on equipment failures. Building on this, a dynamic opportunistic maintenance strategy is developed by integrating the dynamic interaction mechanism of equipment failure rates, capacity allocation ratios, and buffer states on maintenance time windows. The elasticity of time windows is adjusted through a Sigmoid function to optimize maintenance scheduling. Case studies demonstrate that compared to traditional fixed time window models, the proposed model effectively increases the probability of normal system operation and reduces total maintenance costs.