Model input variables are, among others historical measurement data, heat demand forecasts and decision variables of the heat network optimizer.

The optimizer plans the operation of several combined heat and power plants and pumping stations (calculates the optimal parameters of temperature and pressure in the heat source, and optimal pressures for pumping stations in the time horizon of 120 hours).

The optimal planning for the commissioning and shutdown of peak heat sources has also been taken into account.

The solution consists of 56,000 forecast models that predict heat consumption for the entire network, for network areas and for individual users.

In addition, the system uses a simplified model of the district heating network, consisting of 2,000 partial models. Developed using machine learning algorithms, these models predict district heating temperatures, pressures and flows.

Model input variables are, among others historical measurement data, heat demand forecasts and decision variables of the heat network optimizer.

The optimizer plans the operation of several combined heat and power plants and pumping stations (calculates the optimal parameters of temperature and pressure in the heat source, and optimal pressures for pumping stations in the time horizon of 120 hours).

The optimal planning for the commissioning and shutdown of peak heat sources has also been taken into account.

This means optimizing about 1,000 decision variables, taking into account about 20,000 constraints. The system is used by analysts and network dispatchers.

The recommendations proposed by the system allow to reduce CO2 by 16,500 tons per year and reduce heat losses by approx. 5%, which translates into annual savings of PLN 1 million (~250k €).