Pyrolysis parameter based optimization study using response surface methodology and machine learning for potato stalk

Background: The depletion of fossil fuel supplies, along with ever-increasing energy needs, mandates the investigation of clean and renewable fuels. In this regard, the present investigation pursued to assess the suitability of response surface methodology (RSM) and machine learning strategy for optimising the process parameters of potato stalk (PS) pyrolysis. Methods: The experiment was performed in a tubular reactor, and key process factors for example temperature (400 - 650 degrees C), heating rate (50 - 100 degrees C/min), and N2 flow rate (150 - 200 ml/min) were optimised for maximum bio-oil yield. The key features of the produced liquid product (bio-oil) and solid product (biochar) were investigated. Significant Findings: The PS physicochemical study demonstrated enormous bioenergy potential, with higher carbon content (45.82 %), calorific value (17.6 MJ/Kg), and lower moisture content (7.2 wt. %). The coefficient of variation for bio-oil biochar was 1.78 and 1.91 % (less than 10 %), indicating that the model is more reliable and reproducible. The artificial neural network (ANN) better forecasted the process yield; nevertheless, the RSM model successfully forecasted the pyrolysis factors interface and importance. The GCMS analysis of the bio-oil revealed 33.42 % hydrocarbons, 13.42 % esters, 4.62 % acids, 1.71 % ethers, 11.1 % ketones, 14.01 % alcohols, 2.34 % amides, 4.96 % nitrogen-containing substances, and 7.12 % phenols.