arXiv subjects

José Luis Guzmán

Publications and source records attributed to José Luis Guzmán.

7 recordsLinked to original sources

arXiv2026

From Learning to Control: Data-Driven Multi-Agent Reinforcement Learning for Multivariable Control in a Microalgae Bioprocess

Effective control of bioprocesses is particularly challenging due to the intrinsic nonlinearity and dynamic variability of living-cell systems. In microalgae-based photobioreactors (PBRs), maintaining stable pH and dissolved oxygen (DO) levels is critical for optimal growth and productivity, yet their strong coupling and sensitivity to environmental fluctuations make multivariable control difficult. This study proposes a novel hybrid offline-online Multi-Agent Reinforcement Learning (MARL) framework for simultaneous pH and DO regulation, leveraging Deep Deterministic Policy Gradient (DDPG) agents to achieve a fully data-driven and model-free control solution. The agents are trained using historical data generated by an expert system, eliminating the need for direct experimentation with the environment. After deployment, the agents operate autonomously, continuously fine-tuning their policies daily to adapt to evolving process dynamics and reject fast transient disturbances. Experimental validation in an open, industrial-scale PBR at the University of Almeria demonstrated the framework's capability to maintain stable operation under realistic conditions. The results confirm that model-free MARL control provides a robust and adaptive alternative for complex bioprocess environments.

eess.SY↗

arXiv2026

Unlocking feedforward capabilities in Model Predictive Control algorithms to deal with measurable disturbances

Disturbance rejection is a central objective in process control, particularly when measurable disturbances can be exploited through feedforward action. Although Model Predictive Control (MPC) naturally incorporates disturbance models and prediction capabilities, standard formulations cannot achieve complete disturbance rejection since the cost function penalises control effort. This limitation prevents MPC from reproducing the behaviour of classical feedforward compensators. This work proposes a novel framework to embed true feedforward capabilities within MPC without removing the control effort penalty. The approach introduces a dual-control structure in which two control actions are computed simultaneously: a tracking-oriented action addressing set-point tracking and robustness, and a feedforward-oriented action dedicated to disturbance rejection. Both contributions are combined into a single control signal on which the process constraints are explicitly enforced. The feedforward-oriented action is formulated without penalising control effort, enabling full compensation of measurable disturbances. The methodology is developed for Dynamic Matrix Control (DMC), Generalised Predictive Control (GPC), and state-space MPC. Its effectiveness is demonstrated through simulation studies, including comparisons with standard MPC and classical feedforward schemes. A case study based on a reverse osmosis process shows that the proposed approach improves disturbance rejection while preserving constraint handling and overall control performance.

eess.SY↗

arXiv2026

Anti-Windup in PID Control: Review, Analysis, and New Tuning Directions

Actuator saturation is a fundamental nonlinearity that significantly degrades the performance of PID-controlled systems by inducing integrator windup, leading to overshoot, slow recovery, and even instability. Although numerous anti-windup strategies have been proposed, their practical tuning remains largely heuristic and suboptimal in many industrial scenarios. This paper presents a comprehensive comparative study of classical and advanced anti-windup techniques for PI-controlled first-order-plus-dead-time (FOPDT) processes under a wide range of operating conditions. The analysis includes dynamic and instantaneous back-calculation, conditional integration, and adapted schemes. In addition, a novel hybrid anti-windup strategy is proposed, combining conditional integration with dynamic back-calculation to improve responsiveness during saturation, whilst preserving smooth recovery dynamics. Moreover, a key contribution of this work is the development of systematic tuning rules for the tracking time constant in back-calculation schemes, specifically optimised for load-disturbance rejection. These rules are derived from an extensive optimisation study that considers the saturation ratio, controller aggressiveness, and disturbance characteristics. The resulting guidelines provide simple yet effective formulas that achieve near-optimal performance without requiring complex computations. Simulation results demonstrate that the proposed methods significantly outperform commonly used heuristic rules, particularly in disturbance rejection scenarios, and provide clear, practical recommendations for selecting and tuning anti-windup strategies in industrial applications.

eess.SY↗

arXiv2026

Model-Free Control approach for pH Regulation in Thin-Layer Photobioreactors

Thin-layer photobioreactors (TLRs) exhibit fast hydrodynamic and thermal dynamics, strong nonlinear photosynthetic responses and significant time-variability due to irradiance fluctuations and biomass growth. These characteristics challenge conventional model-based control strategies, whose tuning degrades under rapidly changing operating conditions. This work presents the experimental implementation of a model-free control approach, Extremum Seeking Control (ESC), for performance optimization in a semi-industrial thin-layer photobioreactor. Unlike previous studies in raceway ponds, the reduced hydraulic inertia of TLR systems enables the adaptation of this control strategy to accelerate convergence while preserving gradient estimation accuracy. The proposed approach is experimentally compared against classical on-off control and ESC configurations with and without feedforward compensation of solar irradiance. Beyond control performance metrics, biological indicators such as biomass concentration and productivity are evaluated to assess the impact on process efficiency. Results show that the proposed ESC strategy reduced cumulative CO$_2$ consumption by approximately 39 % and decreased the accumulated pH tracking error by more than 60 % compared with conventional on-off control, while biomass- and irradiance-normalised indicators confirmed a more efficient use of injected carbon. These results demonstrate that high-frequency ESC can improve regulation performance and carbon utilisation efficiency in fast photobioreactor systems, highlighting its suitability for thin-layer cultivation under outdoor conditions.

eess.SY↗

arXiv2026

Sensor-Based Turbidostat Operation Enables Biomass Setpoint Regulation and Productivity Improvement in semi-industrial Microalgae Raceway Pond

This work presents the experimental validation of a turbidostat strategy for biomass control in a semi-industrial outdoor raceway reactor. The proposed approach regulates biomass concentration by automatically triggering dilution when the online biomass estimate exceeds a predefined threshold. To ensure safe outdoor operation, dilution was restricted to daylight periods, avoiding biomass removal under low-radiation conditions. The strategy was implemented through an industrial control architecture using an optical monitoring system for online biomass estimation. Experiments were conducted over 14 consecutive days in an 80 m$^2$ (12000 L) raceway reactor. A second parallel reactor operated in chemostat mode, with a nominal dilution of 20 % of the total volume during operating days, provided contextual information under the same outdoor conditions. The analysis focuses on the ability of the sensor-based strategy to configure and maintain the desired biomass concentration, rather than on a direct reactor-to-reactor performance ranking. During the campaign, the biomass threshold in the turbidostat reactor was changed from 1.0 to 0.8 g L$^{-1}$, demonstrating the flexibility enabled by online biomass monitoring. Excluding initial adjustment and transition days, harvested areal productivity increased from 9.52 to 23.20 g m$^{-2}$ d$^{-1}$ after reducing the operating threshold. The overall biomass balance also showed higher net areal productivity in the turbidostat reactor, reaching 20.34 g m$^{-2}$ d$^{-1}$ compared with 11.16 g m$^{-2}$ d$^{-1}$ in the parallel chemostat reactor. These results demonstrate the feasibility of robust turbidostat-based biomass control in large-scale outdoor raceway photobioreactors.

eess.SY↗

arXiv2025

Enhancing industrial microalgae production through Economic Model Predictive Control

The industrial production of microalgae is an important and sustainable process, but its actual competitiveness is closely related to its optimization. The biological nature of the process hinders this task, mainly due to the high nonlinearity of the process along with its changing nature, features that make its modeling, control and optimization remarkably challenging. This paper presents an economic optimization framework aiming to enhance the operation of such systems. An Economic Model Predictive Controller is proposed, centralizing the decision making and achieving the theoretical optimal operation. Different scenarios with changing climate conditions are presented, and a comparison with the typical, non-optimized industrial process operation is established. The obtained results achieve economic optimization and dynamic stability of the process, while providing some insight into the priorities during process operation at industrial level, and justifying the use of optimal controllers over traditional operation.

eess.SY↗

arXiv2025

A Comprehensive Benchmark Platform for Process Control Research of Outdoor Microalgae Raceway Reactors

This paper presents a benchmarking framework to evaluate process control strategies in outdoor microalgae raceway reactors, integrating four key control regulation tasks: pH, dissolved oxygen (DO), culture volume through coordinated harvest-dilution actions, and temperature via a sump-mounted spiral heat exchanger. The benchmark is built upon a high-fidelity, experimentally calibrated dynamic model that captures the strongly coupled thermal, physicochemical, and biological processes governing industrial-scale open raceway ponds. A closed-loop simulation environment is provided, featuring realistic actuator constraints, gas transport delays, stiff integration, and a fully specified scenario based on multi-day outdoor disturbances (irradiance, temperature, wind, and humidity). Four user-replaceable controllers define the manipulation of CO2 injection, air bubbling, harvest/dilution sequencing, and heat-exchanger operation. The platform computes a unified global performance index, in addition to individual metrics for each control problem, combining tracking error, gas and energy usage, and biomass productivity, enabling consistent and quantitative comparison of alternative control strategies. Baseline regulatory architectures (On/Off, PI/PID, and Economic Model Predictive Control (EMPC)) are included to illustrate the benchmark use for classical and advanced control methods. By providing an openly specified, reproducible, and computationally tractable benchmark with well-defined function interfaces, this work aims to bridge control methodology and outdoor algal bioprocess engineering, and to support the development of multivariable control strategies for disturbance-rich environmental systems.

eess.SY↗