Enhancing Kitchen Waste Composting with Black Soldier Fly Larvae: Integrating Life Cycle Assessment and CNN-GRU Models

This study introduces an innovative approach to optimizing Black Soldier Fly Larvae composting by developing a hybrid Convolutional Neural Network-Gated Recurrent Unit model. The model leverages CNN's ability to extract spatial features, such as temperature and humidity, from the composting environment, and combines it with GRU's capacity to process temporal data, capturing the dynamic nature of the composting process. The dataset, sourced from Kaggle, includes detailed food waste data across households, retail, and food service sectors, providing a comprehensive basis for training and testing the model. To ensure consistent model performance, data preprocessing is performed using Min-Max Scaling, which standardizes the data for compatibility with the model’s architecture. The integration of Life Cycle Assessment (LCA) outputs with the CNN-GRU model predictions offers a dual perspective by not only optimizing the composting process but also evaluating the environmental impact, including greenhouse gas emissions and resource use. The proposed CNN-GRU model demonstrates exceptional performance, achieving an accuracy of 98.7%, significantly surpassing traditional algorithms such as Random Forest, Artificial Neural Networks, and standard CNN models. This high level of accuracy underscores the model's effectiveness in predicting and optimizing BSFL composting outcomes. The entire framework has been implemented in Python, enabling robust and flexible deployment in various composting scenarios. This research presents a significant advancement in sustainable waste management practices, providing a powerful tool for optimizing BSFL composting while simultaneously assessing its environmental sustainability. The findings offer valuable insights and practical solutions for enhancing composting efficiency and environmental responsibility.