การประยุกต์ใช้การเรียนรู้ของเครื่องในการพยากรณ์ความต้องการโซ่อุปทานของเศษวัสดุทางการเกษตรสำหรับโรงไฟฟ้าชีวมวล

Abstract

This research emphasizes the importance of knowing the amount of biomass from agricultural residues generated by cultivation activities. It studies and analyzes the components of the biomass supply chain to indicate the readiness of resources that reflect the biomass demand. Therefore, this research forecasts biomass from agricultural residues of five key economic crops and biomass resources for biomass power plants in the northern region, covering 17 provinces in northern Thailand. Historical data related to agricultural production are used as factors in forecasting by applying machine learning models. The ARIMA model is used to forecast leading indicators, and five regression models and one neural network model are employed to determine the most suitable model for the given dataset. The data is divided into 80% for training and 20% for testing. The results include the factors used in the models and the optimal variables for forecasting each crop. Additionally, the forecasted yields are used to calculate agricultural residues and the biomass energy for biomass power plants. It is found that Nakhon Sawan, Kamphaeng Phet, and Phetchabun have the highest amounts of agricultural residues in the northern region, leading to a concentration of biomass power plants in these areas. Thus, knowing agricultural production through forecasting with machine learning applications can reflect the readiness of resources to meet the biomass demand in the northern region in the future.