Abstract:
Energy crisis and emerging negative impacts on environment are the leading factors of industries to increase the share of sustainable resources in the energy production. Biomass based solutions have become as an alternative for fossil fuels due to its availability and sustainability. There are several energy conversion method to utilized biomass, among them. gasification is the one of main energy conversion method. However, biomass gasification has shortcomings due to the barriers like unpredictable variability of biomass properties, process complexity, and controllability of the process.
There are several types of gasification types. Downdraft back bed biomass gasifire is the most suitable one for small power application (10-1000kW) and it is beneficial over the other types because of less complexity of construction and low carbon footprint.
Aim of this study is to develop an artificial neural network based on plant controller for biomass gasifier. Biomass gasification process model was developed using feedforward neural network model (FFNN) and neural network based nonlinear autoregressive model with external output (NNARX). According to results, NNARX showed the best erformance for prediction of process output.The effectiveness of the neural network based internal model controller (IMC) was successfully tested for gasification plant. Two experiments were carried out using 12kg of coconut shells. One experiment plant was run with proposed neural network internal model controller (NNIMC) while second experiment was done without NNIMC and blower was operated at constant reference RPM.
Developed NNIMC was tested using 15kW pack bed imbert type downdraft biomass gasifire and controller algorithms. Performance of introduced IMC was analysed using 72 minutes of continuous plant operation. The analysis revealed that 12% of gasification efficiency can be improved while increased the performance in terms of stability by the introduced of NNIMC.
