A cascade hybrid PSO feed-forward neural network model of a biomass gasification plant for covering the energy demand in an AC microgrid

Agriculture and forestry crop residues represent more than half of the world's residual biomass; these residues turn into synthesis gas (syngas) and are used for power generation. Including Syngas Gensets into hybrid renewable microgrids for electricity generation is an interesting alternative, especially for rural communities where forest and agricultural waste are abundant. However, energy demand is not constant throughout the day. The variations in the energy demand provoke changes in both gasification plant efficiency and biomass consumption. This paper presents an Artificial Neural Network (ANN) based model hybridized with a Particle Swarm Optimization (PSO) algorithm for a Biomass Gasification Plant (BGP) that allows estimating the amount of biomass needed to produce the required syngas to meet the energy demand. The proposed model is compared with two traditional models of ANNs: Feed Forward Back Propagation (FF-BP) and Cascade Forward Propagation (CF-P). ANNs are trained in MATLAB software using a set of historical real data from a BGP located in the Distributed Energy Resources Laboratory of the Universitat Polite`cnica de Vale`ncia in Spain. The model performance is validated using the Mean Squared Error (MSE) and linear regression analysis. The results show that the proposed model performs 23.2% better in terms of MSE than de other models. The tunning parameters of the optimal PSO algorithm for this application were found. Finally, the model was validated to predict the necessary biomass and syngas to cover the energy demand.

Automated summary

This study proposes a model based on Artificial Neural Networks and Particle Swarm Optimization algorithm to estimate the biomass needed for a Biomass Gasification Plant to produce syngas to meet energy demand. The results show that the proposed model outperforms existing models in terms of MSE.