<p>Long time series forecasting is essential in domains such as traffic flow analysis, energy management, and financial modeling. Despite the progress in this area, existing methodologies encounter significant challenges in addressing the elevated computational complexity and excessive memory requirements associated with large-scale time series data, thereby constraining their applicability to long-term forecasting tasks. This study proposes QNNformer, a quantum-enhanced model that integrates quantum neural networks(QNNs) with a quantum multi-attention mechanism to address these limitations. By leveraging quantum computing principles, including parallelism, superposition, and interference, QNNformer effectively captures complex nonlinear relationships and extended temporal dependencies. The model incorporates advanced quantum optimization techniques, including the quantum approximate optimization algorithm and grover diffusion operations, to enhance its capacity for long-term dependency modeling. Empirical evaluations demonstrate that QNNformer achieves state-of-the-art performance in various time series prediction tasks, significantly improving prediction accuracy, computational efficiency, and robustness when handling high-dimensional, noisy, and complex time series datasets. Extensive experiments on seven real-world datasets demonstrate that QNNformer achieves the state-of-the-art performance, outperforming the best baseline by an average of 3.9% and 3.0% in MSE and MAE, respectively.</p>