Influenza poses a persistent public health threat in China, with substantial impacts on health and the economy, especially during seasonal epidemics and emerging outbreaks. Seasonality, local clustering, and serial correlation inherent in influenza data introduce spatio-temporal complexities that traditional statistical process control (SPC) methods cannot adequately capture. This study introduces a novel nonparametric framework for real-time influenza monitoring across 300+ Chinese cities from 2014 to 2020. Reference periods are selected to establish baseline incidence patterns and fit a nonparametric spatio-temporal model to estimate mean and covariance structures. These estimates enable the setting of dynamic outbreak thresholds. Next, exponentially weighted spatial LASSO (EWSL) charting statistics are computed for the monitoring period, prioritizing recent observations and detecting subtle mean shifts in small, clustered regions - well-suited to influenza´s progression dynamics. Charting statistics exceeding control limits trigger timely outbreak warnings. Results demonstrate that our method consistently outperforms alternative methods, and existing literature corroborates that its early signals correspond to actual outbreaks - including those for H7N9 strains, influenza A and B viruses, and the initial spread of COVID-19. These findings highlight the potential of our approach as an effective epidemic monitoring tool, addressing complex spatio-temporal patterns and supporting timely, data-driven public health interventions.