Model Optimization for High-Yield Biocrude in Co-Hydrothermal Liquefaction of Municipal Sludge

With accelerating global urbanization and industrialization, the continuous expansion of urban population size and industrial production capacity leads to increasingly severe challenges in municipal sludge (MS) treatment. Traditional disposal methods such as landfill and composting face bottlenecks including land resource constraints, greenhouse gas emissions, and heavy metal contamination, while incineration suffers from technical issues such as high energy consumption and dioxin pollution. In this context, developing sludge valorization technologies that are both environmentally friendly and economically feasible has become a critical research focus. Hydrothermal liquefaction (HTL) technology is regarded as one of the most promising sludge treatment technologies due to its ability to directly process biomass with high moisture content (80% - 90%). This study innovatively adopts the co-HTL strategy that combines municipal sludge and microalgae, achieving synergy through a feedstock formulation strategy. This approach increases biocrude yields, improves product quality, and reduces the cost of biomass HTL technology, thus facilitating industrial-scale application. The Box-Behnken Design (BBD) was used to develop a three-factor, three-level response surface model, selecting reaction temperature (280 - 340 ℃), residence time (15 - 45 min), and biomass-to-water mass ratio (1∶5 - 1∶15) as key variables. Through 29 sets of experiments, the influence mechanisms of process parameters on biocrude yield were systematically investigated. This study introduces a novel dual-model comparative analysis framework, integrating response surface methodology (RSM) and artificial neural network (ANN). The RSM established a predictive model based on a second-order polynomial equation, achieving an R2 value of 0.9833 and demonstrating excellent linear fitting capability. In contrast, the ANN employed a three-layer topological structure (3-node input layer, 10-node hidden layer, and 1-node output layer). After training with the Levenberg-Marquardt algorithm, the model′s R2 significantly improved to 0.9989, demonstrating the superiority of neural networks in modeling nonlinear complex systems. An increase in temperature significantly promotes biomass decomposition, but secondary reactions (condensation/gasification) above 325 ℃ lead to a decline in biocrude yield. Prolonged residence time results in only marginal yield improvement, while excessive residence time under high-temperature conditions tends to induce side reactions. At lower temperatures, a longer residence time is required to ensure complete reactions, whereas at higher temperatures, the optimal residence time is shorter. Increasing the biomass-to-water ratio from 0.08 to 0.25 g/mL enhances yield by 5% - 7%, yet excessively high ratios may reduce intermediate solubility and inhibit oil phase formation. High sludge ratios significantly suppress yield, primarily because the high ash content (59.1%) dilutes the organic biomass concentration. Finally, a genetic algorithm combined with an ANN was used to predict the optimal process conditions for the co-HTL of MS and microalgae, achieving a maximum biocrude yield of 32.2%. This study offers an innovative solution for sludge resource utilization, and its process optimization framework can be applied to other fields involving the collaborative conversion of organic solid wastes.