Abstract
The energy industry has been transitioning from a fossil fuels based economy to a carbon neutral one. Renewable sources of energy are becoming increasingly more in use as we slowly phase out of coal, petroleum, and other fossil fuels. Although a lot of attention has been given to solar and wind, a major source of energy continues to be biomass. While combustion of biomass is practiced all over the world, gasification has been found to be a cleaner technology due to its homogeneous nature of reactions with air. A standard biomass feedstock is wood; its price, however, has been increasing sharply in recent years, making it economically difficult to continue operating plants with wood chips and wood pellets. Hence, low-cost alternative biomass feedstocks for gasification need to be explored to retain the economical edge. The producer gas (PG) from biomass gasification can be used to substitute fossil fuels in a dual fuel engine to generate combined heat and power (CHP). The present study investigates the performance of a small-scale CHP system, comprising a compression-ignition engine operated with diesel, and producer gas obtained from the gasification of forest residues. Forest residues (FR) here refer to the materials left behind on the forest floor as waste after forest harvesting and management activities. These include small branches, twigs, pine needles, and pinecones. The experiments conducted show the effect of forest residues on the operation of a pilot-scale, downdraft gasifier in comparison to standard wood chips. The findings reveal that the FR mixtures have an effect on the flow behaviors of both feedstock and air in the reactor. While the average composition of the producer gas is similar to that of wood chips, the change in reactor flow leads to an unpredictable nature in the equivalence ratio which is a process controlling parameter. This is reflected in gas parameters such as the flow rate, the lower heating value (LHV) trends, and the cold gas efficiency (CGE), which in turn leads to unpredictability in the overall performance of the CHP system, as evident in its electrical and thermal efficiencies, and exhaust gas components like CO, CO2, and NOx. Diesel substitution rates (DSR) of up to 59% and 46% for partial and full electrical loads respectively were realized. While emissions of CO and CO2 increased with increasing DSR, supplementing diesel with PG has benefits in viii terms of the pollutants NOx and PM, which were found to be generally lower for FR mixtures. In the case of NOx, reductions of over 65% was realized. To address the unpredictability in the PG characteristics and flow, an artificial neural network (ANN) model was developed. The model proved to be reasonably accurate in its predictions with an overall accuracy of about 90% on both the training and validation datasets. Based on the initial architecture, the model was further developed by leveraging custom capabilities which resulted not only in improved performance but also provided better insights in to the functioning of the hidden layers, thus providing a deeper understanding of the so called "black box" model. The final accuracy achieved with the custom model was over 95%. Furthermore, some common issues faced by researchers using deep learning algorithms to model gasification systems are briefly discussed, and a detailed explanation of the custom model developed is provided. Apart from predicting the gas quality, the model can be exploited to improve the practical operation of the gasifier in several ways including better process control and predicting anomalies to reduce downtime, thus optimizing and streamlining its operation.