Comprehensive Decision Framework Combining Price Prediction and Production-Planning Models for Strategic Operation of a Petrochemical Industry

Abstract

This paper presents a new decision-making framework for resource and production planning in the petrochemical industry under price uncertainty conditions. The framework consists of three main decision models: price prediction, paper trading, and production planning. Three different prediction models, system dynamics, multiple linear regression, and artificial neural network, were examined and compared for the precise forecasting of the price of the final products as well as naphtha as a raw material. We then developed a new optimization model using a mixed integer linear programming technique to establish the optimal operational strategies with a given inventory capacity. In the framework, the maximum profitability is ensured by solving a series of decision-making problems on the petrochemical plant, including timing and quantity of naphtha purchase and final product sales, inventory management, financial risk management with paper trading, and operation strategy of the involved processes. To illustrate the capability of the proposed model, the resource and production planning problem of a real petrochemical plant that produces six different major products, ethylene, propylene, butadiene, benzene, toluene, and xylene, was examined. As a result, the optimal production planning using the proposed model enabled the improvement of the total sales by 5.50% and the operating profit by 13.8% in comparison with the BAU case.

Materials inventory optimization using various forecasting techniques and purchasing quantity in packaging industry

“…LSTM and RFR models developed in this study have great reputation to predict, yet there is still lack of past research that studies beyond the prediction specially to use the prediction output as input for decision making. One research example predicts time series Naphtha price that could influence the production of petrochemical products (Kwon et al, 2020) such as Ethylene, Propylene, Butadiene, Benzene, Toluene and Xylene to optimize the profit. The prediction factors are historical prices, quantity of demand and supply, price profiles, crude oil and financial statistics taken sourced from May 2009-December 2010 to train three prediction models multi-linear regression, Artificial Neural Network (ANN) and System Dynamic (SD) which result accuracy of 84% to predict January 2011-December 2011 data.…”

“…Data outlier filtration is not implemented (Kwon et al, 2020) because each data point represents the price volatility of commodity grade material (Sen et al, 2023), that naturally occurs due to supply and demand. Crude oil price unit is US dollar per barrel, meanwhile Naphtha and the rest of variables are in US dollar per ton except Local PP, BOPP target and CPP target are in local currency.…”