تخصیص انتشار CO2 در پالایشگاه های نفتی برای فرآورده های نفتی مشترک: یک مدل برنامه ریزی خطی برای کاربرد عملی

The allocation of CO2 emissions associated with petroleum refineries to petroleum joint products is a necessary step in Well-to-Tank analysis in order to evaluate the environmental impacts of individual automotive fuels. Oil refining is essentially a joint production system and due to the complex nature of the process involved, it is very difficult to establish any noncontroversial allocation pattern for oil products. Under certain conditions, however, refinery linear programming models can provide a non-arbitrary additive allocation schema based on the marginal contribution of each oil product to the total CO2 emissions. But in general, these conditions are not satisfied. In this paper, by extending the LP approach to the optimal Simplex tableau, we propose an original two-stage methodology based on the marginal contribution of oil products and the production elasticity of unit processes to provide an additive CO2 allocation scheme. We show that this procedure emerges from the equilibrium behavior of the refinery and is consistent with microeconomic theory. A numerical example is provided.

The question about industrial energy that this paper addresses is the allocation of the refineries’ CO2 emissions to petroleum joint products. This allocation procedure constitutes a necessary step in evaluating the environmental impacts of automotive fuels in Well-to-Tank (WTT) studies. In fact, WTT is the first part of Well-to-Wheel (WTW) studies and consists in assessing the energy consumption and the resultant CO2 emissions along fuel chains from the extraction of feedstock until the delivery of fuels to the vehicle tanks.1 Since the WTT differences among automotive fuels are due exclusively to the refinery component, especial care should be taken on the allocation principle used to assess the contribution of oil products to the refinery energy use and CO2 emissions. Oil refining is a joint production system and due to the complex nature of the process involved and the vast number of joint product outputs that are strongly correlated, it is very difficult to establish any noncontroversial allocation scheme for oil products. In practice, allocation rules used so far for the petroleum-based fuel are traditionally based on two fundamental approaches: physical measures (mass, volume, energy or exergy2 contents, molecular mass or other relevant parameters) or market value (gross sale value) or expected economic gain of individual oil products from a given refinery. Both of these approaches inevitably involve the use of arbitrary allocation rules. Furoholt (1995) and Wang et al. (2004) point out that these allocation rules should be applied at the sub-process level within a refinery and not at the aggregate process level (i.e., the refinery level). This would consist in partitioning the refinery into different process units and then allocating the energy consumption and the resultant emissions from each process unit to the products from these units according to mass, energy content or market value of final and intermediate petroleum products. They show that, contrary to the aggregate process level, tracking energy use and emissions by individual refining process helps reveal some additional energy and emissions associated with certain refinery products that are otherwise overlooked with the refinery-level allocation. Although the “process-level-based method captures process-dependent characteristics of fuel production within a petroleum refinery” (Wang et al., 2004) it is still open to discussion on two points. First, despite the important effort of tracking the energy use and emissions by individual refining process, this approach still suffers from using arbitrary rules at its final step. In this regard, Azapagic and Clift (1999a,b) show how these arbitrary measures break down in joint production industrial systems when they do not reflect the underlying physical causality and lead to flawed results. For instance, in a Life Cycle Assessment study for the Statfjord production platform for production of regular gasoline, Furoholt (1995) reports that, based on the volume criterion, only 0.5% of the total energy use and the resultant emissions is allocated to gasoline, whereas it is 81% based on the energy criterion and 57% based on detailed partitioning. Second, the sub-process allocation approach provides an incomplete picture of the whole system as it ignores the complex interactions, interdependencies and synergies which exist among the refinery oil products and process units. As a consequence, this approach systematically assigns more energy use and CO2 emissions to the oil products that utilize more process units. An illustrative example of this issue is gasoline and diesel which constitute the two main automotive oil products in WTT and WTW analysis. Most of the existing WTT studies overestimate the environmental burdens (energy use and CO2 emissions) of gasoline due to the higher number of gasoline processing units in European refineries (an exception is the WTT report of CONCAWE and EUCAR, 2006). To show that this conclusion could be wrong, let us consider a standard refining scheme where catalytic reforming units convert low-octane naphthas into high-octane gasoline blending components called reformats with hydrogen as a by product. In response to a continuously increasing demand for diesel (at the expense of gasoline), European refineries have expanded the diesel fraction from oil refining beyond its optimum balance with gasoline yield (Kavalov and Petevs, 2004). This imbalance should worsen due to the tightening of oil product sulphur specifications3 which would also cause the reduction of a refinery's overall output. These changes are associated with higher energy use and CO2 emissions at the refinery level (see e.g., CONCAWE, 2000 and CONCAWE, 2005). For cost reasons, it usually happens that catalytic reforming units operate at full capacity not in order to meet the gasoline demand (which is decreasing in Europe) but to meet the increasing hydrogen requirement of the refinery. Therefore, the additional energy use and resultant CO2 emissions of catalytic reforming units should no longer be assigned to the reforming gasoline output (as would be the case in most of the WTT approaches) but must be solely allocated among the products using the output of hydrodesulphurization units (especially diesel and other middle distillates). This point has some pretty interesting implications for some reforms to the taxation of petroleum products in Europe: due to a lower taxation of diesel compared to gasoline (unequal taxation of the two automotive fuels results in diesel costing about one dollar less per gallon in most European countries), diesel cars dominate new vehicle registrations. This has led to an additional growing demand for diesel (by an average of 4% per year) and a further reduction in gasoline consumption (by an average of 2% per year, over the past five years); and, these changes in domestic demand have driven a significant increase in global trade flows for refined products and blend stocks (Houdek, 2005). Encouraging gasoline demand by changing the existing tax policy (in favour of gasoline) and improving the gasoline engine would probably correct the imbalance.4 The economic logic behind these policy reforms would be strengthened if we show that the European fuel market's evolution (in terms of both quantity and quality) has resulted in higher energy use, CO2 emissions and infrastructure costs associated with diesel production as compared to gasoline. More sophisticated proposals to energy consumption and emission allocation have been developed based on the concept of duality in linear programming (LP). In fact, LP is a powerful mathematical tool which is frequently used to represent the complex scenario of production in the refinery. The advantages of this method compared to the process-level-based approach could be summarized in two points. First, the LP model depicts causality between various inputs and outputs in the refinery and allocates the whole energy consumption and resultant CO2 emissions accordingly without having to use any arbitrary measures (Azapagic and Clift, 1998). Second, the information created through the duality in LP takes into account the complex interdependency and synergy effects among the unit processes and petroleum products in the refinery system. These two “well-behaved” characteristics of the LP-based allocations are consistent with the ISO 14041 recommendations5 and therefore relevant for environmental accounting purposes. Under certain assumptions, these allocation coefficients can be also used for fully assigning the refinery's energy use and CO2 emissions to the oil products. In this particular case, the marginal allocation coefficients can be viewed as the average CO2 contribution of oil products and be used in WTT and WTW analysis. In the literature, this approach has been known as the marginal allocation methodology (Azapagic and Clift, 1994, Azapagic and Clift, 1995, Azapagic and Clift, 1998, Azapagic and Clift, 1999a, Azapagic and Clift, 1999b and Babusiaux, 2003). The major disadvantages of LP models, however, is the difficult access to the technical data to researchers outside of refineries. Most of the time, such data are propriety and generally available only through refinery consultants and operators (Wang et al., 2004). In general, in refinery LP models, process capacity, input availability and some types of calibrating and institutional constraints destroy the additivity property of the LP-based allocations. Due to the fact that the value of any basic variable (as well as the objective function value6) is fully distributed among the active constraints of the model, the petroleum product allocation coefficients along, might underestimate or overestimate the total volume of the refinery's CO2 emissions. This handicap is a valid objection to LP as an allocation tool in WTT and WTW studies and might limit its use for problems in which the objective is to assign unambiguously the whole refinery's energy use and the resultant emissions among the oil products. This methodological-oriented paper is aimed to provide an original two-stage procedure, based on LP, to fully allocate the refinery's CO2 emissions among the refinery's petroleum products. We show that this non-arbitrary (re)allocation mechanism is additive, unique and defensible due to its microeconomic justification. The reminder of this paper is organized as follows. Section 2 briefly reviews the issue of allocating the refinery's CO2 emissions via linear programming. Notations are introduced and the need for a reallocation procedure is discussed. Section 3 delineates the suggested (re)allocation methodology. A numerical example is provided to illustrate the procedure. Last section concludes.

In this paper, we tackled a well-known polemical problem of how to allocate the CO2 emissions of a petroleum refinery to its petroleum joint products. In effect, due to the complex interdependency, synergy and interaction effects among the various inputs and outputs, it is very difficult to establish any noncontroversial CO2 allocation scheme based on the traditional approaches used so far in retrospective WTT and WTW studies. The main objective of this paper was to show that how the information associated with an optimal LP solution can be properly used in order to fully allocate the CO2 emissions of a refinery to its joint products, without having to need any further assumptions or information. To this end, we provided an original two-stage procedure, based on the marginal coefficients of the final Simplex tableau, to extract the average allocation coefficients associated with various oil products. These average coefficients include the direct and indirect contribution of each oil product to the refinery's CO2 emissions; and, they depend totally upon the technical and physical relationships that define the operating state of the refinery. Therefore, they are perfectly consistent with the ISO 14041 recommendations and relevant for WTT purposes; especially when the goal of the study is either an efficiency judgment or prescribing policy changes. To the best of our knowledge, these marginal coefficients have never been used in any LP-based allocation methods. A step-by-step numerical example was provided to illustrate the procedure.