Practical Real-Time Data-Driven Approach for District Cooling Plant Operational Optimization

The real estate sector has a dramatic impact on global carbon dioxide emissions, approximately 70% of the emissions are produced by or are due to building operations. With rising global electricity costs and climate change mitigation, there is a significant interest in energy-efficient solutions for building operations. Although comprehensive cooling systems are widespread throughout Europe, most of them lack robust machine learning-based applications in practice. This work presents an application that optimizes in real-time the operation of the district cooling plant based on a machine-learning model. A more than 35% reduction in energy costs is achieved by implementing the model with optimization including hourly electricity market prices, weather forecasts, and estimated demand in buildings. The application is deployed during the cooling season and is driven by real data collected at 15-minute intervals.