Skip to content

Bioreactor ​

  • 3 state
  • 1 control
  • Lagrange cost
  • fixed final time
  • 3 state box
  • 1 control box
  • 3 boundary
  • OptimalControl valid
  • JuMP valid
  • ModelingToolkit valid

The photobioreactor–digester system problem is a benchmark in constrained optimal control. It models the coupled dynamics of a microalgae photobioreactor and an anaerobic digester for methane production. The system includes three state variables: the algae concentration , the substrate concentration , and the biomass concentration . The control variable represents the input flow rate between the two units. The goal is to maximise methane production over a fixed horizon while satisfying biological and operational constraints Bayen et al. 2014.

Mathematical formulation ​

The problem can be stated as

The horizon is fixed to   (rescaled units), corresponding to several day–night cycles.

The model components are:

  • Light model:   , where encodes the periodic day–night cycle,

  • Growth function (Monod law):   .

Qualitative behaviour ​

The optimal solution exploits the periodic light–dark structure: algal growth is favoured during illuminated phases, while methane production dominates during dark phases. The control typically exhibits a bang–bang structure, alternating between minimum and maximum values, with possible singular arcs when substrate and biomass reach balanced levels. The positivity constraint on biomass is usually active at the beginning of the process, shaping the initial control action.

Characteristics ​

  • Nonlinear coupled dynamics with periodic forcing.

  • Bang–bang and singular arc structures in the optimal control.

  • Serves as a benchmark for optimal control methods with periodic and constrained systems.

References ​

  • Bayen, T., Mairet, F., Martinon, P., & Sebbah, M. (2014). Analysis of a periodic optimal control problem connected to microalgae anaerobic digestion. Optimal Control Applications and Methods. doi.org/10.1002/oca.2127 This paper analyses a periodic optimal control problem modelling a coupled microalgae photobioreactor and anaerobic digester. It provides theoretical insights and numerical solutions for maximising methane production under biological and operational constraints.

  • BOCOP examples: Photobioreactor–digester system problem. Examples-BOCOP.pdf This example demonstrates the practical implementation of the photobioreactor–digester system in BOCOP, serving as a benchmark for constrained, nonlinear, periodic optimal control problems.

Parameters ​

ParameterDefault valueUnitMeaning
t00–initial time
tf200–final time, 20 light cycles
β1–coupling of the flow u between the two units
c2–scaling of the methane production in the cost
γ1–conversion of algae into substrate
halfperiod5–half of the period of the light cycle
Ks0.05–half-saturation constant of the Monod growth of the biomass
μ2m0.1–maximum growth rate of the biomass
μbar1–maximum growth rate of the algae, at full light
r0.005–decay rate of the algae
y_l0–lower bound of the algae concentration y
s_l0–lower bound of the substrate concentration s
b_l0.001–lower bound of the biomass concentration b
u_l0–lower bound of the flow u
u_u1–upper bound of the flow u
y_t0_l0.05–lower bound of the initial algae concentration
y_t0_u0.25–upper bound of the initial algae concentration
s_t0_l0.5–lower bound of the initial substrate concentration
s_t0_u5–upper bound of the initial substrate concentration
b_t0_l0.5–lower bound of the initial biomass concentration
b_t0_u3–upper bound of the initial biomass concentration

Every modeling takes other values with the keyword parameters, merged with these: bioreactor(backend; parameters=(t0=…,)).

Reference objective ​

The best objective among the solves that converged, in each modeling, on the default grid of the modeling: the modelings are independent, each has its own reference. See how the references are computed.

ModelingSchemeGridReference objectiveSolvesDistinct solutionsConfigurations agree
JuMP:trapeze600-12.9296902482no
OptimalControl:trapeze600-12.92969024185no

In JuMP with :trapeze, the solves reached 2 distinct solutions: besides the best one, -12.92545174 (gap 3.3e-04, 6 solves). See the results of the modeling for which configurations reached which.

In OptimalControl with :trapeze, the solves reached 5 distinct solutions: besides the best one, -12.92942888 (gap 2.0e-05, 2 solves); -12.92620279 (gap 2.7e-04, 2 solves); -12.92594631 (gap 2.9e-04, 2 solves); -12.92545205 (gap 3.3e-04, 10 solves). See the results of the modeling for which configurations reached which.