Predictive modelling of biomass moisture and availability uncertainty for Co-firing decision-making in the Kenyan tea processing sector

Kenyan tea processing factories largely rely on biomass as their energy source, yet they face inefficiencies due to uncertainties in fuel moisture content, calorific value, and supply. This study developed and validated a stochastic optimization framework to support cost-effective, sustainable biomass fuel-mix decisions. The model applied 800 Monte Carlo simulations to account for variability in the fuelwood moisture content (10–50%) and availability (50–100%), and used a multi-objective genetic algorithm (GA) in fuel-blending optimization. Two modelling regimes were considered: 10–20% moisture content for the mini-boiler validation trials and 12–50% for the factory-based setting. The experimental validation of a 25 kg/h pilot across 21 trials achieved high predictive accuracy for moisture content (R 2 = 0.963) and fuelwood mass (R 2 = 0.955), and moderate predictive accuracy for fuel cost (R 2 = 0. 749). The optimization model attained the lowest production cost of $0.731 at a moisture content of 12%, and fuelwood-fuel share of 100%, though this diverged from typical practice. A balanced solution based on 75% wood availability target yielded a feasible cost of $0.918, harmonizing efficiency with flexibility. Sensitivity analysis established that under wet conditions, costs were primarily driven by moisture content, while under dry conditions, by fuelwood availability. The nonlinear interaction between fuel properties and cost dynamism had been confirmed by the integrated simulation-optimization model, which explained up to 96.9% of cost variance. The validated model is an effective decision-support tool for biomass co-firing in tea factories that allows minimizing costs, improving efficiency and increasing resilience of the fuel supply chain.