ARIMA and Artificial Neural Networks to forecast the CO2 emissions allowances price: application to the design of petrochemical supply chain under uncertainty

Abstract

To face up the threat of global climate change, governments and regulatory agencies are implementing policies to reduce the greenhouse gas emissions. A common climate change policy is to cap CO2 emissions and establishing a price through trading. The idea behind this cap-and-trading scheme is to set a price on carbon emissions and in consequence, a financial incentive to decrease them. After a cap is fixed on emissions, companies are allowed to buy or sell from each other the allowances to emit CO2. Firms exceeding their emissions cap must buy extra credits to cover the excess. Meanwhile, those that do not use all their allowances can sell them, providing the least-polluting firms with an extra revenue. It is expected that the CO2 emissions allowances price increases with time forcing company towards more sustainable technologies. Hence, the key parameter for a successful result is the evolution of that price with time, which is difficult to predict due to the uncertainty associated to it. One approach to predict that trend is to assume that the emissions allowance price exhibits a similar behavior to a share in the stock market. In this work, we analyze the forecasting performance of the autoregressive integrated moving average (ARIMA) and artificial neural networks by comparing with CO2 European Emission Allowances prices reported for previous year. To illustrate the usefulness of this methodology, we design a petrochemical supply chain under uncertainty in the emission allowances price.