This model illustrates some of the features in ABML. It is derived from our volume sample model.

Distributed recursion solver (DR) will work with this model also. You will need an ABML license. To run this model please place the attached dynamically link library, PUBML.DLL, into the folder where pimswin.exe is installed. Re-launch PIMS after placing the PUBML.DLL into the pimswin.exe folder so that PIMS can pick up this dll. PUBML.DLL is a dynamically link library that the user can choose to build to add a custom set of correlations. The PUBML.DLL source code is attached to give the user a feel for what is needed to construct this dll. There is supplemental documentation (not attached) to explain in detail how to build this dll. In this model, a refiner is producing 4 grades of gasoline, URG, UPR, CRG and CR2. The first 2 grades are EPA RFGs and the last 2 are CARB RFGs. All 4 are splash blended with ethanol at the terminal. List of ABML Correlations 1. D86FromPercentOff / D86ToPercentOff to calculate the blended D86 temperature using Hydrocarbon Processing 1994. 2. UBML_RVPF, UBML_RVPR- user defined RVP forward and reverse transforms defined in the PUBML.dll. PUBML.dll is a dynamically link library that the user can create to add custom correlations to ABMLs list of correlations. PIMS does offer a simpler way to add custom correlations through the property calculator or table INDEX (GPRPCALC, GBLNVAL and GNDXR) described in item 5 below. The user however may choose to use PUBML instead when the correlations are too complex for the property calculator, or the correlations need to be protected by a license. The user may also choose to do this because it is a convenient way to package and distribute a standard set of company wide correlations. Note that ABML has a built in RVP/RVP index correlations. The use of PUBML here is for demonstration purpose only. 3. CARBOB2 - calculates the ethanol blended properties needed by the CARB3 correlation. The ethanol value entered in the optional section for CARBOB2 is the default value. The amount of ethanol by blend is specified in table ABMLOPT. Please note that it is in units of volume percent and not volume fraction. Note that ethanol is not a stream nor is the amount of ethanol a decision variable in this model. If this isnt the case, use CARBOB and its set of helper correlations instead. In this model CARBOB2 takes its input from D86FromPercentOff (D86) and UBML_RVPR (RVP). Note that CARBOB2 automatically maps its output into the CARB3 or CARB equation. Mapping in table RFG is not required and in fact ignored. The user however needs to declare a blend to be a CARBOB2 blend in table BLNSPEC. See PIMS documentation on table BLNSPEC. 4. RBOB calculates the ethanol blended properties needed by the EPA RFG emission correlation. This correlation is analogous to CARBOB2 but also calculates the E200 and E300 needed in the EPA RFG emission correlation. In fact RBOB is a combination of CARBOB2 and the E200E300EPA correlations.

5.

6.

7. 8.

The latter correlation is from the EPA RFG emission prediction spreadsheet that converts T50 and T90 to E200 and E300. The ethanol value entered in the optional section for RBOB is the default value. The amount of ethanol by blend is specified in table ABMLOPT. Please note that it is in units of volume percent and not volume fraction. Note that ethanol is not a stream nor is the amount of ethanol a decision variable in this model. In this model RBOB takes its input from D86FromPercentOff (D86) and UBML_RVPR (RVP). Note that RBOB automatically maps its output into the EPA RFG correlation. Mapping in table RFG is not required and in fact ignored. Note that RBOB is valid for EPA RFG and not conventional. It is only valid for the complex equation. The user needs to declare a blend to be a RBOB in table BLNSPEC. See PIMS manual on table BLNSPEC for additional information UBML_DRIVABILITY calculates the drivability index and is defined in PUBML.DLL. The T50 and T90 for EPA RFG come from RBOB. For CARB RFG blends they come from CARBOB2. The mapping of properties by blend is done using table ABMLMAP. In this model we assume that ethanol doesnt affect T10, and T10 comes from D86FromPercentOff. This is not correct. We commit this error here so that we can illustrate the use of table ABMLMAP. ABML has a built in drivability index correlations. The use of PUBML here is for demonstration purpose only. GPRPCALC a user defined 2nd order property entered through the property calculator. 2nd order properties are blend properties calculated from other properties of the blend. GPRPCALC essentially allows the user to declare which of the correlations defined through the property calculator are to be added to ABMLs list of 2nd order correlations. In this model it is used to define a linear version of TV/L. RVP and D86 temperatures for EPA RFG come from RBOB and from CARBOB2 for CARB RFG. Again table ABMLMAP is used to map properties. Note that like the DRIVABILITY INDEX, we assume that ethanol doesnt affect T10 and T10 comes from D86FromPercentOff. ABML has two versions of TVL built in and the use of GPRPCALC is for demonstration purpose only. There is a corresponding version of GPRPCALC for indexes called GNDXR and a corresponding version for blend value called GBLNVAL. Blend value type correlations are like indexes except that the blended index values are the desired blended property. There is no need to perform a reverse transform. This approach to adding custom correlations is simpler than PUBML, but the correlations that can be added this way is limited as to how complex they can be. The user is also limited to 20 correlations for each type. See the PIMS help system for additional limitations. DON Road octane. This is just the average of research and road octane. MTMR/MTMF Mobil transform for predicting research and road octane. It is used to illustrate how to invoke ABML from within a submodel. Note that ABML does not have a build in correlations for the effect of ethanol on octane. The octane spec should therefore be value for the gasoline before ethanol is added.

Blends of interest:

1. URG, UPR are EPA RFG blends 2. CRG and CR2 are CARB RFG blends. Tables of interest: 1. ABML defines the ABML correlations desired, maps the PIMS property tags to ABML properties and set options. The daisy chaining is done by setting the output properties of one correlation as the input properties of another. 2. ABMLOPT defines the CARBOB2 options by blend. 3. ABMLMAP map ABML property tag by blend. The primary purpose of this table is to allow an ABML correlation to take input from several ABML correlations. 4. ABMLSUBF defines the submodels that ABML is required to perform a forward transform. In this model, aromatic, olefin, research and motor octane is used to calculate the indexes, transformed research and motor octane. 5. SCNP, SSFA and SRFT are submodel where ABML will do the forward transformation mentioned in item 4. 6. BLNSPEC - ABML correlations are only activated if you put specs on them. It will automatically activate other ABML correlations that are needed based on the setup in table ABML. 7. PGUESS Specify initial guesses for the ABML correlations. For ABML indexes, supply the properties. For 2nd order correlation provide the guesses of the inputs to the correlation on the lowest rung of the daisy chain. Couple of additional notes: 1. Daisy chaining refers to PIMS ability to use the output of one ABML correlation as the input of another correlation. This daisy chain is set up in table ABML by using the output property tags as the input tags of another ABML correlation. User can set up the ABML correlations in table ABML in any order. PIMS will automatically determine the ordering of the calculation. 2. Admittedly table ABML looks quite intimidating. In the sample model folder there is a subfolder called completeabml. In it you will find a spreadsheet macrocompleteabml.xls. It contains a wizard that will help you construct the skeleton of table ABML. Just fill in the corresponding PIMS tag in the red cells. The optional data are in the yellow cells. You can either fill them in or delete the unused ones. 3. User needs to supply initial guesses for D86 and RVP for the relevant blends. ABML will automatically determine the other properties that need to be recursed and pad table PGUESS with initial guesses for them. The guiding principle as to what needs to be in PGUESS is as follows. If the correlation is an index type, provide the PGUESS of the blended property. This is because the user generally does not know the index transform and thus have no way to guess what the index values are. For 2nd order correlations, just provide the guesses for the inputs to the correlation of the starting rung of the daisy chain. For example, CARBOB2 uses the output of the D86 and RVP correlations so there is no need to provide initial temperature and RVP guesses for CARBOB2. PIMS will invoke D86 and RVP to

calculate the T50, T90 and RVP needed by CARBOB2. However, the user should examine the recursion range to ensure that they are wide enough. This range can become too small if the initial PGUESS are the same for all the blends and there isnt enough variability in the property entered in other tables to broaden this range. 4. Currently using ABML within submodels is some what limited. You can use ABML index correlations to translate properties into index in the ASSAY and BLNPROP tables and feed them to the submodels. The submodel can operate on these indexes and ABML can then pick up these manipulated values and operate on them when calculating the properties of the blend. Take a look at the crude cut LN1 and its RVP and D86 properties to see how this is done. 5. You can also invoke ABML to perform the forward transform (i.e., index and blend value type correlations) within a submodel through table ABMLSUBF. This is useful in the case where a material is being produced in a submodel and its properties needed by the forward transform are recursed. For example, the Mobil transform for octane needs the research and motor octane, aromatic and olefin of the gasolines components. In the alkylation submodel, SSFA, alkylate (ALK) is being produced. Its octane indexes, TRN and TMN are functions of its aromatic (ARO), olefin (OLF), research and motor octane (RON and MON) values. These values are recursed from the c3= and c4= and therefore can not be calculate at the start of recursion like in item 3. It is therefore necessary to build additional recursion structures to perform the calculation for TRN and TMN during the iterations. See PIMS help for more information on using table ABMLSUBF. 6. The percent off in the D86FromPercentOff and D86ToPercentOff are percent off at TBP and not D86 temperatures. Dont confuse them with the likes of E200 and E300. There is another ABML correlation, EVAPORATIVE which does a simple linear interpolation of the blended D86 curve. However, it is just a simple linear interpolation at this time. There is also the E200E300EPA correlation that calculates E200 and E300 from D86 T50 AND T90 for typical RFG gasoline. This correlation can be tuned by blend. 7. Typically sulfur is declared a weight spec in table WSPEC. In this model CARBOB2 also calculates the effective amount of sulfur after the addition of ethanol. This sulfur property should also be declared a WPSEC. Take a look at SUL and SUl in table WSPEC. 8. It is recommended that for models that deal with RFGs, sulfur should be units of ppms rather than in weight percent. This is because the derivative of the emission level can be quite large when using units of weight percent and it is difficult to tighten the tolerance on sulfur sufficiently so that the net error in the emission due to sulfur recursion error is small. 9. The IBP and FBP of the blended D86 depend strongly on the set of temperatures with which we do the D86 to percent off conversion. They may not be reliable. 10. Unlike CARB, EPA does not mandate the use of the WSPA correlation (CARBOB) to predict the effect of ethanol on the finished gasoline. RBOB is a combination of CARBOB along with the E200 and E300 correlation found in the EPA RFG emission prediction spreadsheet. The users should test the prediction

of RBOB. The RVP, E200 and E300 values can be tuned by blend. See the PIMS manual for additional info.