Forecasting working days of weather sensitive manufacturing using the Kalman filter

Abstract

This paper presents an approach to forecasting working days in weather-dependent types of production using the Kalman filter. The application domain covers industries that are directly influenced by meteorological conditions, including construction, agriculture, road maintenance, and logistics – sectors where timely planning of resources and workforce is critically important. A mathematical model is proposed that enables dynamic assessment of workability based on time series of meteorological variables such as air temperature, wind speed, humidity level, and precipitation.The model was implemented in Python using the open NASA POWER API as a source of up-to-date weather data. The Kalman filter was applied to reduce noise in the input data, smooth observation series, and generate more accurate real-time estimates of system state. The resulting forecast automatically generates «workable» or «non-workable» day markers for various locations and work types, allowing enterprises to quickly adapt their production schedules to changing weather conditions. Modeling results confirmed high accuracy and stability of the forecasts, especially in the presence of data gaps or local anomalies. The advantages of the Kalman filter were demonstrated in comparison with threshold-based methods, showing better alignment with actual production calendars. The obtained results confirm the effectiveness of the proposed approach for automated planning in weather-sensitive operations. The solution can be scaled for use in production management IT systems. The model has practical value for enterprises seeking to minimize downtime risks associated with unpredictable weather patterns.