Modeling and control of two stage anaerobic digester for hydrogen production

CHEN4011 Final Project - Semester 2, 2025

Instructions

1. The project is conducted in a group of four. You are free to choose your group.

2. Please notify the instructors of your groups as soon as you form them.

3. If you cannot find a group, please get in touch with your instructor at the earliest.

4. You will need to submit all the files created electronically on blackboard.

5. There should be one submission per group.

6. Please follow the instructions given below carefully for preparing the files for submission. Failure to follow these instructions may result in us not being able to assess the files.

   You will be uploading two files.

   1. Report (pdf file containing the report). You need to name the file as STUDENTID_CHEN4011_project_report.pdf (Replace STUDENTID with your Student ID). You need to make only one submission per group.

   2. Create a zip file named STUDENTID_CHEN4011_project_Supporting_files.zip. The zip file should contain a) All supporting files for modeling, simulation, and control activities (matlab, simulink) b) the PDF report file. You may upload the supporting file to a cloud storage of your preference and share a link.

Problem description

Anaerobic digestion (AD) is an established route for converting organic wastes into biogas. Traditionally, the focus of AD has been methane () production in a single-stage digesters. However, advances in process engineering have shown that separating the process into two distinct stages, a hydrogenic (acidogenic) stage followed by a methanogenic stage, can significantly improve overall energy recovery. In such a two-stage anaerobic digestion (TSAD) systems, the first reactor produces hydrogen and volatile fatty acids, while the second reactor converts these intermediates into methane (Chorukova, Simeonov, and Kabaivanova (2021)). Energy yields in TSAD systems are reported to be 20–40% higher than conventional single-stage configurations, making them an attractive pathway for sustainable energy production from lignocellulosic and agro-industrial wastes.

Despite these advantages, the design and operation of TSAD systems present major challenges. Reactor performance depends strongly on the ratio of working volumes, dilution rates, and substrate loadings. Optimizing these parameters requires reliable dynamic models that capture the nonlinear interactions of microbial growth, substrate conversion, and inhibition effects (Chorukova, Simeonov, and Kabaivanova (2021)). In addition, AD systems are highly sensitive to disturbances and variability in feed composition, necessitating advanced control strategies to maintain stability and maximize hydrogen yield.

Conventional control methods such as SISO PID controllers are simple to implement but often struggle with the nonlinear, multivariable, and disturbance-prone nature of AD systems. The presence of large time delays, slow microbial dynamics, and strong coupling between hydrogenic and methanogenic stages makes process stabilization particularly difficult. Recent research has demonstrated that advanced nonlinear strategies, such as synergetic control, can significantly improve stability and convergence speed compared to PID, particularly for hydrogen production in cascade digesters (Messili et al. (2025)).

However, these advanced strategies rely on the availability of accurate, high-fidelity models of the digestion process. Developing such models is challenging due to the biological complexity and variability of TSAD systems, and machine learning–based modeling approaches offer a promising pathway to provide the required fidelity for predictive control design.

The problem addressed in this project is therefore twofold: (i) to model and analyze a two-stage anaerobic digester for sequential hydrogen and methane production, focusing on the influence of reactor configuration and operating parameters; and (ii) to evaluate control strategies, including machine learning-based predictive methods, for regulating hydrogen output under realistic disturbances and nonlinearities.

Suggested steps

1. Review existing literature on mathematical and data-driven modeling of two-stage anaerobic digestion (TSAD) systems. Prepare a brief literature review highlighting key findings on hydrogenic and methanogenic reactor dynamics, volume ratio optimization, and control strategies.

2. Implement the nonlinear TSAD models presented in Chorukova, Simeonov, and Kabaivanova (2021) and Messili et al. (2025) using MATLAB/Simulink. Validate that the implemented models reproduce key dynamic and static characteristics reported in the literature.

3. Generate simulation datasets of key process variables (substrates, biomass concentrations, hydrogen and methane flow rates) under different dilution rates and substrate loadings.

4. Use these datasets to train a machine learning model (e.g., artificial neural network or recurrent network) capable of predicting hydrogen production rate from process inputs and states. Divide the data into training and testing sets and evaluate predictive performance using suitable metrics (e.g., RMSE, $R^2$).

5. Identify control, manipulated, and disturbance variables for the TSAD process. For example, manipulated variables may include dilution rates, while outputs of interest include hydrogen and methane flow rates. Disturbances may include fluctuations in inlet substrate concentration.

6. Design and evaluate different control strategies, including conventional PID and advanced approaches. Using the machine learning model as a surrogate, develop suitable transfer function approximations (SISO or MIMO) for controller design.

7. Compare controller performance under set-point changes and disturbance scenarios. Assess convergence speed, robustness, and stability across different strategies.

Report

Prepare a report consisting of the following:

1. Literature review

   Present a comprehensive note (5–6 pages) on modeling and control of two-stage anaerobic digestion (TSAD). Discuss published models, reactor configurations, microbial pathways, and control approaches (PID, synergetic, machine learning). Summarize key challenges such as nonlinearities, long delays, and coupling between stages.

2. Modeling and simulation of TSAD

   Implement the nonlinear dynamic models reported in the literature and simulate reactor performance for hydrogen and methane production. Demonstrate key features such as static characteristics, time delays, and substrate inhibition.

3. Data-driven modeling

   Develop and train a machine learning model (e.g., ANN or recurrent network) using simulation data. Evaluate its ability to predict hydrogen production and other key outputs, and compare against the mechanistic model.

4. Control strategy design and evaluation

   Identify manipulated, controlled, and disturbance variables. Design controllers (PID, synergetic, ML-based predictive control) and test their performance under set-point changes and disturbance scenarios. Use appropriate metrics (settling time, ITAE, RMSE) to benchmark results.

5. Critical review.

   Provide a reflective analysis of your modeling and control work: Which aspects were reliable? What assumptions and simplifications were made? How can both the mechanistic and ML-based models be improved for future studies?

Marking

Description	Marks
Literature review	20
Modeling and simulation of TSAD	20
Data-driven modeling	20
Control strategy design and evaluation	20
Critical review	10
Report presentation	10
Total	100

See detailed rubric in Section 6 for marking key.

Report format

The following guidelines are presented to ensure uniformity, clarity, and professionalism in your report submission.

Cover page

The cover page should have the following information

Project Title: Modeling and Control of a Two-Stage Anaerobic Digestion System for Hydrogen Production

Submitted by: Student Names and IDs:

• Student 1 (ID: )
• Student 2 (ID: )
• Student 3 (ID: )
• Student 4 (ID: )

Date of Submission: DD Month YYYY

Peer Contribution

Team Member	Overall Contribution (%)
Member 1
Member 2
Member 3
Member 4

General requirements

1. Maximum Length:
   • 30 pages total (excluding references, appendices, and nomenclature).
   • Pages exceeding this limit will carry a 10% penalty.
2. Font & Text Formatting:
   • Font: Standard professional font either sans or sans serif
   • Font Size: Main text 11 pt minimum; captions, footnotes, references 9–10 pt minimum
   • Line Spacing: 1.15 minimum
   • Text Alignment: Justified
   • Paragraph Spacing: 6 pt after each paragraph
   • Section Headings: Use numbered sections (e.g., 3.2 Modeling approach)
   • Subheadings: Use consistent formatting
3. Page Layout:
   • Paper Size: A4
   • Margins: minimum 2 cm on all sides
   • Header/Footer: May be used for page numbers and project title
   • Page Numbers: Bottom-center or bottom-right, starting after the cover page

Figures, tables, and equations

• Figures/Tables:
  • Must be numbered (e.g., Figure 3.2, Table 5.1)
  • Caption placed below figures, above tables
  • Cite in text (e.g., “as shown in Figure 3.2”)
• Equations:
  • Center aligned
  • Numbered on the right (e.g., (1), (2))
  • Use consistent symbols and define them in the nomenclature section

References

• Citation Style: APA6 or Chicago (consistent throughout)
• All references must be cited in-text
• Include journal articles, books, and relevant technical standards
• Suggested minimum: 10 quality references

Technical writing

• Avoid informal language
• Use passive or formal voice (e.g., “The model was implemented using…”)
• Define all acronyms upon first use
• Be concise and clear – avoid redundant explanations
• Avoid large blocks of text – use figures, tables, and bullet points where suitable

Report structure

The following structure is recommended for your report:

Cover page
1.0 Executive Summary
2.0 Introduction
3.0 Literature Review
    3.1 TSAD process
    3.2 Modeling approaches
    3.3 Control strategies
4.0 Modeling and Simulation
    4.1 Governing equations
    4.2 Assumptions and limits
    4.3 Simulation results
5.0 Data-Driven Modeling
    5.1 Simulation dataset
    5.2 ANN model
    5.3 Model validation
6.0 Control Strategy
    6.1 Process variables
    6.2 Controller design
    6.3 Performance metrics
7.0 Critical Review
8.0 Conclusions
9.0 References
10.0 Appendices

Appendices should include supporting material that is too detailed for the main body but still essential for completeness, transparency, or reproducibility. Examples include:

• Detailed derivations and calculations not shown in the main report
• Complete list of model parameters, coefficients, and input data
• MATLAB/Simulink files or code snippets used for simulation and machine learning model training
• Additional simulation plots and sensitivity analyses

Use section numbers (e.g., Appendix A, Appendix B) and reference them in the main text (e.g., “see Appendix B for parameter values”). Ensure readability and proper formatting even for raw data or code.

Submission

• The project is conducted in a group of four.
• You are free to choose your group.
• Please notify the instructors of your groups as soon as you form them.
• If you cannot find a group, please get in touch with your instructor at the earliest.

Submission instructions:

• You will need to submit all the files electronically on Blackboard.
• There should be one submission per group.
• Please follow the instructions given below carefully for preparing the files for submission. Failure to follow these instructions may result in us not being able to assess the files.

You will be uploading two files:

1. Report (PDF file containing the report). Name the file as: STUDENTID_CHEN4011_project_report.pdf (Replace STUDENTID with your Student ID. Only one submission is required per group.)

2. A zip file named: STUDENTID_CHEN4011_project_supporting_files.zip The zip file should contain:

   1. All supporting files for modeling, simulation, and control activities (MATLAB, Simulink, Python, etc.) used in the project
   2. A copy of the PDF report file

   You may upload the supporting files to a cloud storage platform of your choice and include a shareable link inside the zip file if the size is too large.

Submission checklist

• Cover page includes project title, student names, group number, and date
• All required sections are included in the report
• Page limit not exceeded
• References formatted and cited properly
• Figures and tables clearly labeled
• Equation numbering is consistent
• Proofread for grammar and clarity
• Zip file includes supporting files and PDF report

Marking rubric

Marking Criteria	Unsatisfactory (Fail) [0–49%]	Satisfactory (Pass) [50–59%]	Competent (Credit) [60–69%]	Very Competent (Distinction) [70–79%]	Excellent (High Distinction) [80–100%]
Literature review (20 marks)	No or incomplete review; little or no coverage of modeling or control aspects. [0–9.8]	Basic review with minimal detail on TSAD; limited mention of modeling or control. [10.0–11.8]	Covers key TSAD modeling and control aspects with some technical depth. [12.0–13.8]	Detailed discussion with good structure, comparisons, and literature support. [14.0–15.8]	Thorough, well-structured, and critically informed review with strong referencing and recent advances. [16.0–20.0]
Modeling and simulation of TSAD (20 marks)	No clear model, major errors, or incomplete implementation. [0–9.8]	Basic model setup with limited explanation or results; little validation. [10.0–11.8]	Clear model development based on literature with some analysis of outputs. [12.0–13.8]	Detailed simulation using Chorukova model with validation of static/dynamic features. [14.0–15.8]	Complete, well-validated implementation capturing nonlinearities, delays, and inhibition, with thorough justification. [16.0–20.0]
Data-driven modeling (20 marks)	No attempt at ML modeling or serious methodological errors. [0–9.8]	Simple ML model attempted with limited dataset use and evaluation. [10.0–11.8]	ML model trained on simulation data with reasonable predictive performance. [12.0–13.8]	Robust ANN/RNN implementation with systematic training/testing and analysis. [14.0–15.8]	Excellent use of data-driven modeling with strong validation, comparison to mechanistic models, and insightful interpretation. [16.0–20.0]
Control strategy design and evaluation (20 marks)	No clear control problem identified; incorrect or incomplete controller setup. [0–9.8]	Basic controller attempted (e.g., simple PID) with limited testing. [10.0–11.8]	Controllers designed and evaluated with some performance benchmarking. [12.0–13.8]	Multiple strategies (PID, synergetic, ML-based) tested with meaningful comparisons. [14.0–15.8]	Comprehensive control design with systematic evaluation, clear benchmarking (settling time, ITAE, RMSE), and strong insights. [16.0–20.0]
Critical review (10 marks)	No reflection or irrelevant content. [0–4.9]	Basic reflection on results and limitations. [5.0–5.9]	Identifies strengths and weaknesses with some suggestions. [6.0–6.9]	Thoughtful insights on reliability and improvements with clear reasoning. [7.0–7.9]	Structured, critical review with benchmarking or literature comparison and clear improvement pathways. [8.0–10.0]
Report presentation (10 marks)	Poorly structured, major formatting/grammar issues. [0–4.9]	Basic structure and formatting, some clarity issues. [5.0–5.9]	Clear layout with acceptable writing and referencing. [6.0–6.9]	Professional formatting and writing with minimal errors. [7.0–7.9]	Polished, professional report, free of errors, in the tone of a graduate engineer. [8.0–10.0]

References

Chorukova, Elena, Ivan Simeonov, and Lyudmila Kabaivanova. 2021. “Volumes Ratio Optimization in a Cascade Anaerobic Digestion System Producing Hydrogen and Methane.” Ecological Chemistry and Engineering S 28 (2): 183–200. https://doi.org/10.2478/eces-2021-0014.

Messili, O., S. Semcheddine, A. Chaabna, and University of Ferhat Abbas Setif. 2025. “Synergetic Control of Cascade-Configured Anaerobic Digestion Bioreactors for Enhanced Hydrogen Production.” Revista Mexicana de Ingenierı́a Quı́mica 24 (2): 1–10. https://doi.org/10.24275/rmiq/IE25556.