Research on Supply and Demand Forecasting and Energy-Saving Scheduling in Smart Canteens

Abstract

This paper presents an integrated solution for a smart canteen in a large industrial park, encompassing supply and demand forecasting, meal preparation and procurement, and energy-saving scheduling. Based on operational and energy consumption data from the past four years (2018-2021), a two-tiered demand forecasting model (daily/hourly) was developed. Furthermore, under time-of-use pricing and peak-limit restrictions, coordinated improvements were implemented to kitchen and HVAC equipment. Results show that in the 2025 test dataset, the MAPE value of the random forest forecasting model was 3.31%, a 68.1% reduction compared to the empirical lag method. Regarding meal preparation, the leftover rate decreased from 9.05% to 3.36%, resulting in a total annual saving of approximately 9.45 × 10⁻⁶ kJ/m³. ^One ton of food (calculated at 0.35 kg/meal) saved approximately 1.4 million yuan in raw material costs. From an energy consumption perspective, predictive model control reduced annual electricity consumption by 9.0%, electricity costs by 13.2%, average daily peak power by 14.9%, and carbon emissions from electricity by 38.9 tCO2. This research can provide engineering references for public institutions to coordinate "anti-waste" and "energy conservation and carbon reduction".