Reaching Efficient & high-Load Operation of Anaerobic Digestion – RELOAD

RELOAD's originality lies in coupling two scientific fields: biology and hydrodynamics (the study of fluid flow). Biological models (such as ADM1) usually assume the reactor is perfectly mixed, which is not representative for industrial digesters processing thick mixtures. On the other hand, fluid flow models (CFD) are difficult to couple with biology.
The project first studied in detail the viscosity (rheology) of real mixtures from a partner digester (PAPE). This data were fed into CFD simulations to precisely map the flow and the mixing zones within the industrial reactor.
Concurrently, the ADM1 biological model was adapted and validated using operational data from the same plant.
The central innovation was to «compartmentalize« the reactor: the total volume was divided into interconnected perfectly mixed zones, with flows calculated by CFD. The biological model was then applied to each compartment. This approach allows for simulating spatial heterogeneity and predicting hotspots where acidity accumulates, something a classic model cannot see.
Finally, this coupled model, too complex for real-time use, was simplified using deep learning to create an ultra-fast reduced-order model suitable for process control. It was used as a tool to optimize and control the process.

The project developed a «numerical simulator« of the digester.
Simulations revealed that the flow inside the digester is laminar and the agitators mix 10% of the total volume. The coupled model showed strong heterogeneity: 58% of biomethane production occurs in only 27% of the volume, creating high inhibition risk zones that are invisible in standard models.
Deep learning was used to reduce the computation time of the hydro-biological coupled model, while maintaining excellent predictive capacity. This tool allowed the optimization of operating strategies, potentially increasing biomethane production by 11%.

The RELOAD project has developed a numerical simulation tool for monitoring, controlling, and improving industrial anaerobic digesters. Its originality lies in an approach combining multi-scale experimentation and modeling, with the PAPE digester as a case study.

Among the project's major achievements, the digester downscaling strategy enabled rigorous validation of hydrodynamic simulations by measuring local flows of a model non-Newtonian fluid. Another major achievement is the drastic reduction in model complexity and computation time through deep learning, making it possible to use computationally intensive research models in real-time industrial control systems.

Simulations indicate that a 60% reduction in mixing frequency does not adversely affect hydraulic efficiency, making a 20% reduction in agitation energy very achievable. The 3% increase in yield and the simulation of load increases remain to be evaluated.

Future work could include the integration of multiphase simulations (liquid-solid-gas) to overcome the limitations of single-phase rheological modeling, taking into account local viscosity. An improvement to the CFD-CM model would be the automation of compartment determination and a better understanding of flows. Other optimizations could integrate the simulation of the site as a whole (post-digester, circulation, energy consumption) and the study of robustness in the face of disturbances.

The methodology developed can serve as a model for other anaerobic systems. The coupled CM-ADM1 model is a powerful platform for in silico experimentation. The integration of the rapid learning model into the control system will enable real-time optimization of methane production.

The project's work has led to two doctoral theses and 14 scientific publications submitted to international journals and conferences. These articles focus on the validation methodology for flow models, hydro-biological coupling, and the application of deep learning for model reduction. The developed methodology provides a foundation for future innovations in bioprocess control.


The RELOAD project is a collaborative-enterprise research project (PRCE) coordinated by Air Liquide. It brings together three academic partners: LBE (INRAE), CentraleSupélec (L2S & LGPM), and LRGP (CNRS, Université de Lorraine). The project began in November 2021 for a duration of 48 months. It received ANR funding of €550,034 for a total global cost of approximately €1.63 million.

Anaerobic digestion (AD) generates biogas, from which we extract biomethane, a biofuel which contributes to clean mobility. Support policies to the biomethane sector in France, in particular feed-in tariffs, will decrease in the coming years. So, it is key to intensify biomethane production and decrease production costs, in order to continue to deploy this renewable energy source.

The most widespread digesters are CSTR (Continuous Stirred Tank Reactor) digesters, operated with a dry matter of maximum 10-15%. Mixing plays a key part in the process, that is, homogenization, to maximize productivity and limit sedimentation. However, mixing must not be too strong, to avoid microorganism stress and limit AD energy consumption.

RELOAD aims at developing a comprehensive scientific AD model, coupling biology and hydrodynamics. The model will then be reduced and integrated to carry out remote control and optimize the process. Various goals are set: optimize AD, regarding mixing, piloting and increase the solids content, up to 20%. The process will thus be intensified and production costs will be reduced. Developing a new digital tool in RELOAD is in line with the ADEME roadmap (2017). It corresponds to a need in research, development and innovation to address bottlenecks of the biomethane sector.

The work will be carried out using cereal silage and manure, which are two representative substrates of the French deposit. All scales will be studied, from lab scale to pilot scale, and finally full scale at an industrial site. This project is innovative since no complex coupled hydro-biological models exist today to pilot AD, even less so at high solids levels.

The consortium has complementary fields of expertise, with LBE (expert in AD), CentraleSupélec (CS, expert in process engineering and control command), LRGP (expert in bioreactor processes) and Air Liquide (AL, leading gas company). The project construction is: WP1 (lead AL) - coordination, WP2 (lead LRGP) - hydrodynamics modelling, WP3 (lead LBE) - hydrodynamic model coupling with biological model, WP4 (lead AL) - pilot and full scale validation, WP5 (lead CS) - model application: control command and optimization (innovative impeller, optimized operating parameters). A grant of 550 k€ is requested to the ANR, for a total project cost of 1.6 m€.

This complementary public-private partnership will allow to transfer fundamental biological and physical knowledge to piloting and optimizing full-scale biogas plant operations. The final model will be easily transferable to different industrial sites, by adjusting certain input parameters (rheology, digester configuration, substrate characteristics). The originality is to couple biological and hydrodynamic models, at high solids contents. RELOAD is an ambitious project in line with the ANR Bioéconomie axis.

Project benefits are numerous: (i) scientific, with the development of fundamental models and piloting of complex bioprocesses; (ii) industrial, with the transfer of the tool to multiple AD sites; (iii) economic, with an immediate potential gain of 2.1 m€/year and a market for the monitoring tool licence of 6.1 to 8.1 m€/year in France by 2028, according to an internal study of Air Liquide; (iv) societal, with contribution to the biomethane sector, to develop this renewable energy source.