arXiv:2601.10189v1 Announce Type: new Abstract: Decarbonizing the global energy supply requires more efficient heating and cooling systems. Model predictive control enhances the operation of cooling and heating systems but depends on accurate system models, often based on control volumes. We present an automated framework including time discretization to generate model predictive controllers for such models. To ensure scalability, a primal decomposition exploiting the model structure is applied. The approach is validated on an underground heating system with varying numbers of states, demonstrating the primal decomposition’s advantage regarding scalability.