Publications

• Data-Driven Modeling, Order Reduction and Control of Anaerobic Digestion Processes

  Published in: University of Tlemcen, Jan 2024

  Abstract: This research delves into the realms of data-driven modeling, order reduction, and control strategies within the context of Anaerobic Digestion (AD) processes. The study is centered on addressing pivotal challenges in this domain and delivering innovative contributions to the field. The primary objectives encompass streamlining the complexity of the Anaerobic Digestion Model No.1 (ADM1) for the specific purpose of control, as well as the exploration of suitable data-driven techniques to achieve precise modeling and prediction of AD systems. Furthermore, the research endeavors to extract kinetics reactions from simulated time-series AD data, develop robust predictive models for Chemostat dynamics under both Monod and Haldane kinetics through data-driven methodologies, and employ the Koopman Operator theory to enable data-driven modeling and control of the Chemostat system, relying solely on substrate measurements. By adopting a data-driven approach, this research aims to provide profound insights into the intricacies of AD processes, thereby shedding light on their complex dynamics and advancing our comprehension beyond conventional models. It introduces an alternative modeling perspective exclusively grounded in data, augmenting our analytical capabilities within the realm of AD processes. The research rigorously evaluates and tests a variety of data-driven techniques, yielding intriguing results. Notably, the application of the Koopman Operator theory represents a significant contribution, particularly in scenarios where measurement resources are limited. This innovation holds the potential to pave the way for robust control strategies within AD systems, ultimately enhancing their sustainability and efficiency.

• Data-Driven Control of the Chemostat Using the Koopman Operator Theory

  Published in: U.P.B. Sci. Bull., Series C, 2023

  Abstract: The chemostat is widely used as a laboratory pilot for bioprocess studies. Chemostat models are nonlinear and rarely used in modern control experiments. For a data-driven control strategy, we use the Koopman operator approach to derive a linear model for a simple chemostat with one substrate and one biomass, using only the chemostat’s input-output data. For chemostat control, we use the linear Koopman model to develop a MPC controller. The linear Koopman model best fits chemostat data compared to the local linearization-based model. In addition, the MPC based on the Koopman model gives very satisfying results compard with a linear MPC controller when applied to control the chemostat. The results are gained for a large space of initial conditions when chemostat control is usually limited.

• Modeling of biogas production from hydrothermal carbonization products in a continuous anaerobic digester

  Published in: ACM Trans. Model. Comput. Simul., Just Accepted (July 2024)

  Abstract: The coupling between anaerobic digestion and hydrothermal carbonization (HTC) is a promising alternative for sustainable energy production. This study presents a dynamic model tailored for a lab-scale anaerobic digester operating on HTC products, specifically hydrochar and HTC liquor derived from sewage and agro-industrial digestate. Leveraging a modified version of the Anaerobic Model 2 (AM2), our simplified model of four states integrates pH and biomass decay rates into biomass kinetics. Simulation results of the mode were compared with experimental data collected over 164 days from the digester. The obtained results have proven the ability of the proposed model to predict the trend of the biogas production as well as important measured outputs of the bioreactor. The developed model could be used to control and optimize the performance of the digester, which provides potential for bioenergy production from waste streams such as digestate and digestate treated through the HTC process.

• Data-Driven Determination of the Governing Equations of the Continuously Stirred Tank Reactor

  Published in: 2023 10th ICEEE, Istanbul, Turkey, May 8-10, 2023

  Abstract: This paper presents a data-driven approach for identifying the governing equations or ODEs of a continuously stirred tank reactor (CSTR) system. The paper employs the sparse identification of nonlinear dynamics (SINDy) algorithm, a popular and versatile method used for discovering nonlinear dynamical system models from data. The SINDy-PI (parallel, implicit) framework, a robust variation of the SINDy method, is used to find the implicit dynamics and rational nonlinearities of the CSTR for both Monod and Haldane types of specific growth rates. The simulation results demonstrate the accuracy of the method in inferring the ODEs of the CSTRs using limited and noisy data. The proposed method can be used to improve our understanding of complex systems and inform the design of control strategies.

• Dynamic Mode Decomposition with Control for Data-driven Modeling of Anaerobic Digestion Process

  Published in: 2022 16th African Conference on Research in Computer Science and Applied Mathematics (CARI), Tunis, Tunisia, 2022

  Abstract: This paper proposes a data-driven modeling approach for complex Anaerobic Digestion (AD) systems. This method is called Dynamic Mode Decomposition with Control (DMDc), which is an emerging equation-free technique for deducing global linear state-space input-output models with actuation for complex systems. DMDc is applied to a set of data generated from simulating the Ordinary Differential Equations (ODEs) of the Anaerobic Model 2 (AM2) using MATLAB. The simulation results demonstrate the prediction accuracy of the linear state-space model generated from the DMDc algorithm.

• Forecast of chemostat dynamics using data-driven approach

  Published in: 2021 International Conference on Control, Automation and Diagnosis (ICCAD), Grenoble, France, 2021, pp. 1-6

  Abstract: This paper deals with the forecast of chemostat dynamics using a data-driven approach. We construct a data-driven model (predictor) based on the Koopman operator theory, which can predict the future state of the nonlinear dynamical system of the chemostat by only measuring the input and output of the system. We are presenting a predictor with a linear structure, that can be used for diagnostics, state estimation and future state prediction and control of nonlinear chemostat. Importantly, the method of generating such linear predictors is entirely data-driven and extremely simple, leading to nonlinear data transformation (embedding), and a linear least squares problem in the embedded space which can be readily solved for large data sets. We show in simulations that Koopman approach best predicts the system trajectories compared to a local linearization methods.

• Model order reduction using Dynamic Mode Decomposition: Application to the Anaerobic Digestion Model N.1 (ADM1)

  Published in: 2020 15th African Conference on Research in Computer Science and Applied Mathematics (CARI), Thiès, Senegal, 2020

  Abstract: The Anaerobic Digestion Model No.1 (ADM1) is by far the most detailed model for the simulation and monitoring of Anaerobic Digestion (AD) processes. However, the ADM1 model is not dedicated for control purposes, due to its high dimension with 35 state variables. Dynamic Mode Decomposition (DMD) technique was applied to reduce the ADM1 order, using data generated from the Benchmark Simulation Model No. 2 (BSM2). The method allows to obtain a global linear model with only 7 state variables, which are coherent with dominant dynamics of the ADM1. We show in simulation that we can reconstruct original state variables of ADM1 model.