運用隨機規劃於農產品產銷決策之分析

Abstract

本論文探討國內芒果產銷決策，將收成量及市場需求之不確定性納入考慮，透過數學規劃模型分析，以期協助農民獲得最大的收益。針對收成量及需求量的不確定性，本研究運用巴斯擴散模型，利用各類芒果歷史交易資料進行配適，以取得各期預測量，結合預測誤差建構收成及需求量之機率分佈。本研究個案分析，假設收成及需求皆為不確定性，建構隨機規劃模型，求解採收、儲存、催熟、運輸及銷售之最佳決策，並設計一表格化介面，將複雜的求解結果轉化為易於解讀資訊，提供果農完整的產銷決策。本研究提供由生產到銷售之整合決策方案，降低過去依據經驗法則做決策的風險，為農產品供應鏈研究注入新力。

This study focuses on harvesting and sales decisions for a mango farmer under uncertain yield and demand. We collect the mango sales data in the spot market in Taiwan and apply the Bass diffusion model to obtain a time-series demand. The test data has been applied to validate the forecast model and to recover forecast error distributions. A stochastic programming model is then proposed with the integration of time-series forecasts and random errors. We develop a procedure to solve the stochastic program based on Monte Carlo methods. The approximate problem solution provides the information of picking, inventory, ripening, delivering, and sales of mangos during the harvesting season. We demonstrate the robustness of the proposed approach by comparing deterministic and stochastic solutions. The case study results are summarized as an easy-to-read calendar for a broader decision maker.