Data-driven uncertainty propagation for stochastic predictive control of multi-energy systems

Stochastic predictive control schemes that account for epistemic and aleatoric uncertainties, i.e. lack of model knowledge and stochastic disturbances, are of major interest for multi-energy systems. However, there exists a trade-off between model complexity, computational effort, and accuracy of uncertainty quantification. This paper attempts to assess this trade-off by comparing a recently proposed approach combining Willems’ fundamental lemma with polynomial chaos expansion to a model-based scheme that first propagates uncertainty with PCE and then considers chance constraints in the optimization. The simulation results show that the data-driven scheme yields similar performance and computational efficiency compared to the model-based scheme, with the advantage of avoiding the construction of explicit models.