191 results
Now showing1 - 10 of 191
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Towards grid-aware multi-period flexibility aggregation — a constrained zonotope approach(Elsevier, 2027-07-10); ; ; ; Aggregation schemes provide a means to reduce the computational complexity of power system operation by reducing the number of devices that are considered individually. This can be achieved with tools of computational geometry, where the feasible set is projected onto the decision variables of the point of interconnection. Set projection is computationally expensive, especially in the context of multi-period power system operation. This calls for efficiency improvements via structure exploitation of set representations. This paper proposes efficient flexibility aggregation via constrained zonotopes. We evaluate the performance of the proposed method on a 15-bus distribution grid with time-dependent elements for up to 96 timesteps. The results suggest that the presented method significantly improves computation times compared to classic polytope projection approaches.Publicationtype: Journal Article8 4 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , CO₂-driven pH control in the enzymatic hydrolysis of urea: stochastic model predictive control under uncertainty(American Chemical Society (ACS), 2026-07-09); ; ; Enzymatic urea hydrolysis is a mild route to ammonia production, but urease activity is strongly pH dependent. In buffer-free operation, pH can drift substantially because ammonia and inorganic carbon rapidly repartition through acid−base equilibria. Detailed process-modeling background, experimental protocol, and parameter identification context are reported in a companion manuscript by Dittmer, K. R. (2026). Here, in contrast, we develop an uncertainty-aware modeling and control framework for CO₂-driven pH regulation during urease-catalyzed urea hydrolysis. A control-oriented mechanistic model couples pHdependent urease kinetics with substrate and product inhibition, fast speciation with electroneutrality-based pH computation, and gas−liquid mass transfer of NH3 and CO₂ in a headspace. Parametric uncertainty in kinetic and transfer parameters is propagated using nonintrusive polynomial chaos expansion. Building on these predictions, we propose a stochastic model predictive control (SMPC) scheme that tracks a pH reference under input constraints while enforcing chance constraints on pH safety bounds. Simulations demonstrate uncertainty-aware pH regulation without buffer salts under the considered parametric uncertainty scenarios, providing a foundation for automated operation of enzymatic reactors under uncertainty.Publicationtype: Journal Article49 47 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Turnpikes in Deep Learning: beyond ResNets and neural ODEs?(IEEE, 2026-07); ; ; ; It is well known that deep learning, and in particular the training of ResNets and neural ODEs, can be formalized and analyzed from an optimal control perspective. In this work, we extend the dynamic systems and optimal control perspectives to fully connected neural networks with ReLU activations and no skip connections. By exploiting equivalence relations between ResNets and networks without skip connections, we show that the corresponding training problems exhibit the turnpike property under conditions analogous to those established for ResNets. We illustrate our findings with numerical experiments for the two-spirals dataset and MNIST.Publicationtype: Conference Paper13 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Fast risk certification of candidate trajectories under uncertain time-varying constraints(2026-06-30); ; ; This paper studies the certification of a fixed candidate trajectory on a finite certification grid under parametric uncertainty. For each constraint-time pair, we define a scalar measure of constraint violation and aggregate the resulting pointwise chance constraints into a worst-case Value-at-Risk (VaR) margin. The goal is not to generate a new trajectory, but to assess online whether a trajectory produced by a planner or predictive controller is sufficiently safe on the certification grid. Direct evaluation requires repeated uncertainty propagation and is often too expensive for computationally demanding models. We therefore adopt an offline-online scheme: offline, a surrogate of the constraint violation map along the candidate trajectory is constructed using polynomial chaos expansion (PCE) when the uncertainty law is known, or kernel regression when only sampled input-output data are available; online, the surrogate is sampled to evaluate conservative VaR bounds at low computational cost. On the theoretical side, we derive a finite-sample upper bound for the grid-based VaR margin using empirical quantiles, the Dvoretzky-Kiefer-Wolfowitz (DKW) inequality, and a union bound over all constraint-time pairs, without assuming a parametric family for the underlying violation distribution. We also show how a uniform surrogate error bound transfers to the certified VaR margin. The approach is illustrated on a crystallization population balance model, where the surrogate-based risk estimates track direct Monte Carlo results while substantially reducing online evaluation time.Publicationtype: Preprint51 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Uncertainty propagation under residual disturbances: a smart-home case study(2026-05-15); 22 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Data-driven continuous-time optimal control: a unified framework using orthogonal functionsWe study data-driven optimal control of continuous-time linear systems over finite- and infinite-time horizons. Our approach builds on our continuous-time version of Willems et al.’s fundamental lemma and on the use of orthogonal basis functions to approximate system trajectories. We show that the solution to an optimal control problem can be approximated by a finite linear combination of basis functions and we establish error bounds for such approximations. Moreover, we approximately solve the algebraic Riccati equation and the associated optimal controller gain directly from data, opening up the possibility of optimal controller design directly from data analogue devices.Publicationtype: Journal Article12 20 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , On dissipativity of cross-entropy loss in training ResNets — A turnpike towards architecture searchThe training of ResNets and neural ODEs can be formulated and analyzed from the perspective of optimal control. This paper proposes a dissipative formulation of the training of ResNets and neural ODEs for classification problems. Specifically, we consider a variant of the cross-entropy (label smoothing) as a loss function and as a regularization in the stage cost. Based on our dissipative formulation of the training, we prove that the training OCPs for ResNets and neural ODEs alike exhibit the turnpike phenomenon. We illustrate this finding with numerical results for the two spirals and MNIST datasets. Crucially, our training formulation ensures that the transformation of the data from input to output is achieved in the first layers. In the following layers, which constitute the turnpike, the data remains at an equilibrium state and therefore these layers do not contribute to the transformation learned. In principle, these layers can be pruned after training, resulting in a network with only the necessary number of layers thus simplifying tuning of hyperparameters.Publicationtype: Journal Article33 17 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Decentralized real-time iterations for distributed NMPCThis article presents a Real-Time Iteration (RTI) scheme for distributed Nonlinear Model Predictive Control (NMPC). The scheme transfers the well-known RTI approach, a key enabler for many industrial real-time NMPC implementations, to the setting of cooperative distributed control. At each sampling instant, one outer iteration of a bi-level decentralized Sequential Quadratic Programming (dSQP) method is applied to a centralized optimal control problem. This ensures that real-time requirements are met and it facilitates cooperation between subsystems. Combining novel dSQP convergence results with RTI stability guarantees, we prove local exponential stability under standard assumptions on the MPC design with and without terminal constraints. The proposed scheme only requires neighbor-to-neighbor communication and avoids a central coordinator. A numerical example with coupled inverted pendulums demonstrates the efficacy of the approach.Publicationtype: Journal Article42 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Reducing power consumption of blowers in emergency ventilation by field weakeningField-weakening is an established technique for exceeding motor base speeds while maintaining constant input power. To improve energy efficiency of emergency ventilators, this paper presents a methodology for reducing torque to lower the input power of brushless DC (BLDC) motors during phases of constant pressure. Using the Clarke-Park transformation, the operating point is obtained by minimizing the voltage and then applied as phase-advance angle and modulation index in the voltage model. A proof-of-concept experimental setup yields power savings of up to 40% using field-weakening compared to conventional field-oriented control approaches, demonstrating its feasibility for energy savings in BLDC motors.Publicationtype: Conference Paper20 12 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Data-driven modeling with prior system knowledgeThe behavior of a linear time-invariant system can be characterized entirely by measured input–output data that spans the vector space of all possible trajectories of the system relying on the fundamental lemma by Willems et al. However, useful a priori knowledge of the system is usually neglected. We propose a novel method for incorporating prior knowledge, specifically, known pole and zero locations, into a data-driven representation by constructing filters that pre-process the measured input–output data. To this end, a physics-informed data-driven predictor is introduced, where trajectories are obtained as linear combinations of the columns of a filtered block-Hankel matrix. We explicitly derive the output prediction error and show how leveraging prior knowledge reduces the impact of future noise realizations on output predictions and improves the accuracy of the initial state that is inferred from past data.Publicationtype: Journal Article26 30