District heating systems (DHS) are today the most efficient and cost-effective method for distributing heat to residential and commercial buildings in urban areas. In 2021 DHS market size in Europe was 131 MUSD, with residential users’ uptake of 59.2% . In Serbia, overall heat production in DHS in 2022 was 6662 GWh, with natural gas as the main fuel (73%) . The costs of fuel only for these systems were as high as 200 MEUR.
Even though DHS are based on mature and proven technology, there are tremendous opportunities for improvement of their operation, especially related to decreasing the operation costs and reducing their carbon footprint. One approach to increased DHS effectiveness is utilization of waste heat, particularly from combined heat and power (CHP) plants. While profitable on the longer-term, this approach requires significant capital investments. Another approach is related to exploiting significant unlocked potential for reengineering of present short- and long-term operation strategies of DHS. Implementation of new, intelligent control schemes based on recent advances in the Artificial Intelligence (AI) field could be a game-changer, providing considerable savings with minimum investment.
The general objective of the project is to improve the short- and long-term operation strategy of DH systems through development of novel control solutions based on explainable AI which will provide instant and measurable economic, environmental and societal benefits.
AI can facilitate accurate forecasting of the actual heat load by the consumers and optimal control parameters and real-time detection of sub-optimal system operation. Those capabilities are essential for optimized heat production and consequently, minimized fuel consumption, waste and CO2 emissions reduction, maximum consumer satisfaction and more effective short and long-term heat production planning.
The key specific objectives of the project are to:
- SO1. Improve accuracy of short- and long-term DHS plant operations planning by using predictive modeling techniques based on modern AI architectures, and thus minimize consumption of resources and CO2 emissions.
Indicators: overheating occurrence rate during the trial period in relation to the rate during the project duration (target is 10%), end users thermal comfort satisfaction (more than 80% of survey respondents) - SO2. Facilitate transparent and inclusive process of heat consumption in urban areas with Explainable AI models and tools enabling synthetic data insights, explanations for heat fluid supply variations, quantifying impact of different factors on room temperatures and what-if analyses, by the citizens.
Indicators: Number of active end-users (target is 30 active users of XAI4HEAT app), end-users’ satisfaction (more than 80% of survey respondents) - SO3. Achieve capability of real-time detection and explanation of problems (anomalies) leading to increased consumption of resources, and negative environmental effects.
Indicators: Number of detected anomalies acted upon in relation to the overall number of detected anomalies (target is 90%), control response time for detected anomalies acted upon (target is 20 min)
Project vision is to achieve a reduction of minimum 10% in heat production in the small 8MW DHS system which will be used for experimentation. This saving will be achieved by reducing the overheating arising in statically controlled DHS, by implementing dynamic control that directly correspond to actual heat demand in a real time and to some extent, by raising awareness of the end users on the impact of their behavior to the actual heat production. Mapped to the economic and environmental benefits, 10% reduction corresponds to annual savings of 40.000 EUR and 200 t less CO2 emissions. When those savings are scaled to the typical sizes of urban DHS systems (for example, 250MW for City of Niš, Serbia), they become quite significant.