Francois Pacaud

Biography

François Pacaud is a researcher specializing in the optimization of energy systems and large-scale infrastructure networks. His work focuses on developing advanced numerical methods to solve complex problems in nonlinear optimization, particularly in the fields of electricity markets, microgrids, and dynamic systems under uncertainty. He explores innovative approaches such as gradient-based iterative methods, primal decomposition algorithms, and the use of parallel architectures (GPUs) to accelerate the solution of large-scale problems. His research also incorporates stochastic optimal control techniques, such as stochastic dual dynamic programming (SDDP) and model predictive control (MPC), applied to the optimal management of energy networks and distributed systems. The evolution of his work reflects a growing expertise in adapting optimization solvers to high-performance computing architectures, while ensuring scalable and robust solutions for industrial and academic applications.

Publication(s)

2025
Strategic bidding in energy markets with gradient-based iterative methods DOI : 10.1016/j.segan.2025.101878
2025
Scalable Primal Decomposition Schemes for Large-Scale Infrastructure Networks DOI : 10.1109/TCNS.2025.3526709
2025
LEVERAGING GPU BATCHING FOR SCALABLE NONLINEAR PROGRAMMING THROUGH MASSIVE LAGRANGIAN DECOMPOSITION DOI : 10.1137/21M1450112
2024
Accelerating Condensed Interior-Point Methods on SIMD/GPU Architectures DOI : 10.1007/s10957-022-02129-5
2024
Optimization of a domestic microgrid equipped with solar panel and battery: Model Predictive Control and Stochastic Dual Dynamic Programming approaches DOI : 10.1007/s12667-022-00522-7
2024
Accelerating optimal power flow with GPUs: SIMD abstraction of nonlinear programs and condensed-space interior-point methods DOI : 10.1016/j.epsr.2024.110651
2024
GPU-accelerated dynamic nonlinear optimization with ExaModels and MadNLP DOI : 10.1109/CDC56724.2024.10886720
2024
Parallel interior-point solver for block-structured nonlinear programs on SIMD/GPU architectures DOI : 10.1080/10556788.2024.2329646
2023
Constrained Policy Optimization for Stochastic Optimal Control under Nonstationary Uncertainties DOI : 10.23919/ACC55779.2023.10156553
2023
Exploiting GPU/SIMD Architectures for Solving Linear-Quadratic MPC Problems DOI : 10.23919/ACC55779.2023.10155791
2022
A feasible reduced space method for real-time optimal power flow DOI : 10.1016/j.epsr.2022.108268
2022
Distributed Multistage Optimization of Large-Scale Microgrids under Stochasticity DOI : 10.1109/TPWRS.2021.3087775
2021
Domain Decomposition Preconditioners for Unstructured Network Problems in Parallel Vector Architectures DOI : 10.1145/3458744.3473363
2020
Mixed Spatial and Temporal Decompositions for Large-Scale Multistage Stochastic Optimization Problems DOI : 10.1007/s10957-020-01733-7
2020
Exact converging bounds for stochastic dual dynamic programming via fenchel duality DOI : 10.1137/19M1258876
2018
Stochastic decomposition applied to large-scale hydro valleys management DOI : 10.1016/j.ejor.2018.05.025

Teaching

Optimization

Lecturer

This optimization course covers finite-dimensional convex nonlinear optimization. It begins by presenting the fundamentals of convex analysis and the sufficient and necessary conditions for optimality. Next, optimization algorithms—first for unconstrained problems and then for constrained problems—are discussed in detail, along with detailed proofs of convergence guarantees. Finally, elements of advanced convex analysis are covered to provide an introduction to non-smooth optimization methods. This course includes lectures (10 hours), tutorials (8 hours), and practical sessions conducted in Python (6 hours). Detailed ContentChapter 1: Optimality Conditions and Convex Analysis (3 hours)Definitions, optimality conditions, convex analysis (convex function, subdifferential, optimality conditions, strong convexity)Chapter 2: Numerical Methods for Differentiable Optimization (15 hours)2.1 Unconstrained Optimization: Gradient Methods (optimal step, linear search, stochastic gradient)Newton and quasi-Newton (BFGS)2.2 Constrained OptimizationLagrange multipliers, stationarity conditionsKKT conditions, active constraint algorithmsDuality and saddle points, Uzawa’s algorithmChapter 3: Introduction to Non-Smooth Optimization (6 hours) Advanced convex analysis: Fechner transform, proximal operator; subgradient methods, proximal gradient method, bundle methods