Industrial processes, crucial for economic progress, generate emissions and pollutants. Smokestacks, acting as both ventilation and emission outlets, play a vital role in effective production. However, the resulting environmental consequences necessitate advanced monitoring. State-of-the-art technologies, such as Chimney Automatic Monitoring Devices or Tele-Monitoring Systems (TMS), instigate a shift. Real-time TMS data provides insights into emission dynamics and temporal patterns. Acknowledging that achieving optimal control requires more than current data, this study explores the prediction of future emissions. The research asserts that advanced knowledge of emissions empowers adaptive control, including the implementation of a soft sensor. Utilizing predictive models like the Transformer-based soft sensor enables proactive adjustments, ensuring alignment with regulations and sustainable practices. The study introduces a forecasting system based on transformers, with a focus on comprehending emission dynamics through soft sensor integration. The aim is to anticipate environmental changes rather than simply exerting control. Specifically, the research employs the Transformer model to forecast pollutant levels using the soft sensor, overcoming limitations in capturing dependencies. The dataset focuses on Total Suspended Particles (TSP), gathered from a TMS within a high-purity FeSI plant. The Transformer model, coupled with the integrated soft sensor, captures intricate patterns in time series data. The methodology encompasses preprocessing TSP data and configuring the Transformer model with the soft sensor to predict pollutant concentrations. By exploiting the self-attention mechanism, the model anticipates complex temporal dependencies, demonstrating superior performance. This research aligns with the development of a cost-effective, cloud-based smart factory platform for air quality management, contributing to sustainable industrial practices. In its early stages, the study holds promise for providing innovative insights into the application of Transformer models with soft sensors in environmental time series forecasting, advancing predictive analytics for air quality monitoring in manufacturing.