Prediction of biochar characteristics and optimization of pyrolysis process by response surface methodology combined with artificial neural network

Improving the biomass pyrolysis process can effectively improve the physical and chemical properties of biochar, making soybean straw biochar return to the field more effective. In this study, using the combination of response surface method (RSM) and Artificial Neural Network of Non-dominated Sorting Genetic Algorithm-II (ANN-NSGA-II), taking the biomass mesh number, heating rate, and residence time as influencing factors, the effects of pyrolysis conditions on relative the specific surface area (SSA), electrical conductivity (EC), and pH value were studied systematically, and the pyrolysis process was optimized. The combination of RSM and ANN-NSGA-II solves the problem that multiple responses cannot be optimized simultaneously in the same RSM model because of different response center values. Based on the results, the heating rate has the greatest impact on biochar SSA. The biomass mesh number has the greatest impact on biochar EC. The biomass mesh number and residence time have great impact on biochar pH. The prediction performance of ANN-NSGA-II is higher than that of RSM. Through the ANN-NSGA-II, the optimized pyrolysis process is that the soybean straw biomass mesh number is 89, the heating rate is 6.3 celcius/min, and the residence time is 115 min. The SSA of soybean straw biochar obtained by pyrolysis is 114 m2/g and EC is 7.8337 ms/cm and pH is 10.5. RSM combined with ANN-NSGA-II can effectively explore the impact of pyrolysis conditions on physical and chemical properties, predict the physical and chemical properties of biochar, and achieve multi-objective optimization.