The refinery hydrogen distribution system usually comprises a set of hydrogen main headers (pipelines) working at different pressures and hydrogen purities. Many makeup and recycle compressors drive the hydrogen through this complex network of consumer units, on-purpose production units, and platformers (see Figure 1a). On-purpose hydrogen plants generate high purity hydrogen at different costs while net production units are platformers generating low purity hydrogen as a byproduct. Hydrogen streams with different purities, pressures and flow rates coming from make-up hydrogen plants and platformers are supplied to multiple consumer units through the hydrogen main headers. Purge streams from hydrotreaters containing non-reacted hydrogen are partially recycled and mixed with fresh hydrogen streams from hydrogen headers before re-routing them to consuming units .
The remaining off-gas stream is burnt as fuel gas. By controlling the fuel gas flow, the purity of the recycled hydrogen stream can be adjusted (Figure 1b). The major hydrotreater operating constraint is a minimum hydrogen/hydrocarbon ratio along the reactor in order to avoid carbon deposition over the catalyst and its premature deactivation. As the catalyst cost is very significant, an effective operation of the hydrogen network will help to increase the catalyst run length, thus boosting the refinery profitability. Moreover, some consuming units may have group of membranes that can be activated to separate and recycle higher-purity hydrogen streams to the hydrogen piping network (Figure 1b).
The MINLP mathematical model
The integrated management of the whole refinery hydrogen network is a very challenging task that requires effective computer-aided optimization tools. The key principle behind the hydrogen management is the fact that not all processes need hydrogen of the same purity. This section describes the proposed MINLP framework for the cost-effective management of refinery hydrogen systems. Main model decision variables and constraints permit to write accurate hydrogen mass balances in terms of purity and flowrate for every stream. The model aims at systematically improving the use of existing refinery hydrogen supplies as a network problem. Its main goal is to minimize the hydrogen production cost while satisfying predefined hydrocarbon production targets, actual topological and operational restrictions as well as minimum utility hydrogen needs at desulphurization reactors. Problem constraints related to hydrogen production units, headers and consumer units are introduced below.
1. Hydrogen production unit constraints. As previously stated
, a refinery system usually
comprises several production units, i.e. H2-plants and catalytic reformers,
that can simultaneously be supplying hydrogen streams with different levels of
purity and pressure to the pipeline network. Therefore, if an existing
production unit uÎPU is being operated
in the refinery, i.e. Yu = 1, equations (1)
and (2) will enforce the corresponding lower and upper limits on hydrogen
flowrate (Qu) and purity (Pu), respectively.
However, it is worth mentioning that hydrogen streams generated by platformers as
a byproduct usually have a certain flowrate and purity, and consequently they
become model parameters. Here, it should be noted that the optimization model
will be able to choose the most convenient operating conditions for the
alternative hydrogen sources in order to meet hydrogen demands at minimum cost.
Equation (3) defines the amount of hydrogen feed that is being directly
supplied from production units to alternative hydrogen headers hÎH and consumer units uÎCU.
2. Hydrogen pipeline constraints. The refinery pipeline network receives high-purity hydrogen streams coming from producer units and medium/low-purity streams from platformers and consumer unit recoveries. Different headers are usually operated at a given hydrogen purity and partial pressure. Equations (4) and (5) enforce a hydrogen mass balance between inlet and outlet streams in every header. Therefore, if at a given moment the hydrogen production exceeds the actual consumption, the balance is satisfied by supplying the surplus hydrogen to the refinery fuel gas system. In turn, equation (6) computes the header hydrogen purity (Ph) taking into account the total hydrogen flowrate in the header (Qh), the flowrate of hydrogen inlet streams coming from alternative sources (quh) and their corresponding purities (Pu and Poutu).
3. Hydrogen consumer unit constraints. Consumer units carry out different hydrotreating operations by utilizing the hydrogen streams available in the network. Equation (7) computes the total hydrogen feed (Qinu) being supplied to consumer unit u from different sources while the bilinear equation (8) determines the actual purity (Pinu) of the combined hydrogen inlet stream. In turn, equation (9) forces a minimum purity requirement for the combined inlet stream of every consumer unit. The minimum hydrogen need for processing the oil fraction (cu) being treated in unit u is specified by equation (10) by enforcing a minimum hydrocarbon/hydrogen ratio. Equations (11) and (12) predict the flowrate (Qoutu) and purity (Poutu) of the non-reacted hydrogen stream from unit u. These estimations are obtained by using non-linear correlations fq and fp that are functions of the flowrate and purity of the inlet streams as well as the inherent features of the oil fraction being hydrotreated in the unit, i.e. density, sulphur and aromatics content, etc. Finally, equation (13) determines the amount of off-gas that is being recycled and supplied to headers and other consumer units.
4. Objective function. The proposed objective function computes the total hydrogen cost required for hydrotreating pre-specified oil-fractions. The non-linear correlation fc calculates the total production cost as a function of the current hydrogen purity and flowrate in each producer unit u. This function may easily accommodate internal and/or external hydrogen suppliers with different cost and restrictions. Alternatively, the proposed model with minor changes could be used for maximizing the refinery profitability. In this case, the model may optimally select the oil-fractions to be hydroteated subject to minimum and maximum oil-fraction demands and a maximum hydrogen availability. This scenario seems to be particularly interesting for dealing with ultra low-sulphur targets and, consequently, future hydrogen shortfalls.
A case study of a H2 network comprising two on-purpose plants, two platformers and eight hydrotreating units with different needs of hydrogen purity and flowrates is depicted in Figure 2a. In turn, Figure 2b shows the optimal hydrogen balance when the HD3 hydrogen purity need decreased to 95.9%. The optimal balance generated by the MINLP model with modest CPU time obtained a 25% reduction in H2 production cost.
Conclusions and future work
An MINLP-based approach has been presented to optimally manage complex hydrogen networks of refinery operations. The proposed model is able to systematically reduce utility cost by increasing hydrogen recovery in consumer units and reducing production cost in the alternative hydrogen suppliers. This project stage is mainly focused on a rigorous treatment of hydrogen mass balances. Future work will aim at extending the model to also consider actual compression costs and operational restrictions as well as the use of alternative separation units (membranes) to recycle higher-purity off-gas to consumer units.