Gas Emission Characterization and Monitoring Algorithm in the Process of Agricultural Waste Resource Treatment

Abstract

To ensure that there is enough food produced for a growing population, trends in population growth and food consumption must be monitored. Therefore, farmers adopt various strategies without taking into account their detrimental effects on the surroundings to fulfil the necessary requirements. These kinds of actions frequently result in higher emissions of greenhouse gases . The use of fertilizer, soil, animals, and other factors all contribute to GHGs  in the agriculture sector. In this Manuscript, Gas emission characterization and monitoring algorithm in the process of agricultural waste resource treatment (GECMA-AWRT-RDCNN)is proposed. Initially, the data is collected from Agri-food CO2 emission dataset. Then, the collected data is fed into pre-processing utilizing Implicit Bulk‑Surface Filtering (IBSF). The IBSF is used for data cleaning. Then the preprocessed data undergoes feature selection process. Here, it selects 14 features by utilizing High Level Target Navigation Pigeon Inspired Optimization (HLTNPIO). Then the selected features are given to Robust Deformed Convolutional Neural Network (RDCNN) for predicting Gas Yield (GY) generated by energy recovery from agricultural waste. RDCNN often does not explain techniques for optimising parameters through adaptation. Hence, the Fractional Order Water Flow Optimizer(FOWFO)to optimize Robust Deformed Convolutional Neural Network which accurately predict the Gas Yield. The proposed GECMA-AWRT-RDCNN approach is implemented in Python. The proposed method’s performance examined utilizing performance metrics likes Accuracy, Mean-squared error (MSE), Root mean squared error (RMSE), Mean bias error (MBE), Determination coefficient , Relative root mean squared error (rRMSE), ,Mean absolute percentage error (MAPE) and Loss. The proposed GECMA-AWRT-RDCNN approach contains 28.0%, 27.5%, and 26.5% higher accuracy, 26.0%, 23.5%, and 28.5% higher Determination Coefficient and 12.0%, 17.5%, and 16.5% lower Mean Squared Error compared with existing methods, such as Electricity production based forecasting of greenhouse gas emissions in Turkey with deep learning, support vector machine and artificial neural network algorithms (EP-GHGE-SVM) Prediction of Agricultural Emissions in Malaysia Using Machine Learning Algorithms (PAEM-ARIMA)  and Role of deep learning for prediction of greenhouse gas emission from agriculture: enabling technology (PGHGA-LSTM), respectively.